AVX2 Support for key-value sorting

[sterrettm2]

This patch adds support for AVX2 key-value sorting. For comparison, benchmarks comparing the AVX2 vs. AVX512 performance and AVX2 vs. scalar performance are below:

AVX2

--------------------------------------------------------------------------
Benchmark                                Time             CPU   Iterations
--------------------------------------------------------------------------
simdkvsort/random_128/uint64_t        1939 ns         1946 ns       360240
simdkvsort/random_1k/uint64_t        18021 ns        18029 ns        39110
simdkvsort/random_100k/uint64_t    2359494 ns      2359426 ns          298
simdkvsort/random_1m/uint64_t     27181339 ns     27180247 ns           26
simdkvsort/random_10m/uint64_t   328366429 ns    328336094 ns            2
simdkvsort/random_100m/uint64_t 3818560389 ns   3818247990 ns            1
simdkvsort/random_128/int64_t         1939 ns         1948 ns       359936
simdkvsort/random_1k/int64_t         16188 ns        16201 ns        43224
simdkvsort/random_100k/int64_t     2094296 ns      2094317 ns          335
simdkvsort/random_1m/int64_t      24546170 ns     24543133 ns           28
simdkvsort/random_10m/int64_t    300429155 ns    300408697 ns            2
simdkvsort/random_100m/int64_t  3539816011 ns   3539431827 ns            1
simdkvsort/random_128/double          1657 ns         1665 ns       420794
simdkvsort/random_1k/double          13263 ns        13268 ns        52672
simdkvsort/random_100k/double      1707007 ns      1706945 ns          411
simdkvsort/random_1m/double       20710996 ns     20709242 ns           34
simdkvsort/random_10m/double     264312414 ns    264284784 ns            3
simdkvsort/random_100m/double   3124498665 ns   3124159045 ns            1
simdkvsort/random_128/uint32_t        1117 ns         1120 ns       626058
simdkvsort/random_1k/uint32_t         6418 ns         6428 ns       109035
simdkvsort/random_100k/uint32_t     884360 ns       884394 ns          791
simdkvsort/random_1m/uint32_t     10349431 ns     10348265 ns           68
simdkvsort/random_10m/uint32_t   129223761 ns    129208665 ns            5
simdkvsort/random_100m/uint32_t 1529426690 ns   1529322850 ns            1
simdkvsort/random_128/int32_t         1117 ns         1120 ns       625407
simdkvsort/random_1k/int32_t          6465 ns         6473 ns       108458
simdkvsort/random_100k/int32_t      922079 ns       922080 ns          758
simdkvsort/random_1m/int32_t      10840572 ns     10839740 ns           65
simdkvsort/random_10m/int32_t    134945071 ns    134931366 ns            5
simdkvsort/random_100m/int32_t  1591200513 ns   1591027400 ns            1
simdkvsort/random_128/float           1205 ns         1209 ns       578252
simdkvsort/random_1k/float            7378 ns         7390 ns        94558
simdkvsort/random_100k/float        975399 ns       975467 ns          716
simdkvsort/random_1m/float        11259949 ns     11258612 ns           62
simdkvsort/random_10m/float      138409369 ns    138397578 ns            5
simdkvsort/random_100m/float    1634287725 ns   1634146491 ns            1

AVX512

--------------------------------------------------------------------------
Benchmark                                Time             CPU   Iterations
--------------------------------------------------------------------------
simdkvsort/random_128/uint64_t        1174 ns         1177 ns       575703
simdkvsort/random_1k/uint64_t         7565 ns         7572 ns        92599
simdkvsort/random_100k/uint64_t    1099925 ns      1100012 ns          637
simdkvsort/random_1m/uint64_t     14643507 ns     14642890 ns           48
simdkvsort/random_10m/uint64_t   199687140 ns    199670122 ns            4
simdkvsort/random_100m/uint64_t 2487791770 ns   2487510844 ns            1
simdkvsort/random_128/int64_t         1177 ns         1180 ns       593556
simdkvsort/random_1k/int64_t          7617 ns         7625 ns        92009
simdkvsort/random_100k/int64_t     1099157 ns      1099172 ns          638
simdkvsort/random_1m/int64_t      14583034 ns     14581973 ns           48
simdkvsort/random_10m/int64_t    199769132 ns    199758441 ns            4
simdkvsort/random_100m/int64_t  2486518465 ns   2486306230 ns            1
simdkvsort/random_128/double          1079 ns         1081 ns       643342
simdkvsort/random_1k/double           6218 ns         6225 ns       112379
simdkvsort/random_100k/double      1005576 ns      1005649 ns          694
simdkvsort/random_1m/double       13543725 ns     13542388 ns           52
simdkvsort/random_10m/double     189044901 ns    189014084 ns            4
simdkvsort/random_100m/double   2369940116 ns   2369678581 ns            1
simdkvsort/random_128/uint32_t         878 ns          880 ns       795229
simdkvsort/random_1k/uint32_t         3586 ns         3588 ns       195088
simdkvsort/random_100k/uint32_t     548641 ns       548656 ns         1278
simdkvsort/random_1m/uint32_t      6862367 ns      6862099 ns          102
simdkvsort/random_10m/uint32_t    95574488 ns     95569892 ns            7
simdkvsort/random_100m/uint32_t 1191578120 ns   1191506814 ns            1
simdkvsort/random_128/int32_t          877 ns          879 ns       796114
simdkvsort/random_1k/int32_t          3591 ns         3594 ns       194537
simdkvsort/random_100k/int32_t      545403 ns       545420 ns         1272
simdkvsort/random_1m/int32_t       6858498 ns      6857780 ns          103
simdkvsort/random_10m/int32_t     95449697 ns     95438473 ns            7
simdkvsort/random_100m/int32_t  1190093444 ns   1190017700 ns            1
simdkvsort/random_128/float            942 ns          944 ns       744034
simdkvsort/random_1k/float            4677 ns         4679 ns       149366
simdkvsort/random_100k/float        622690 ns       622691 ns         1125
simdkvsort/random_1m/float         7654461 ns      7654027 ns           91
simdkvsort/random_10m/float      102934181 ns    102924527 ns            7
simdkvsort/random_100m/float    1270152278 ns   1269981042 ns            1

AVX2 vs. Scalar

Benchmark                                                                        Time             CPU      Time Old      Time New       CPU Old       CPU New
-------------------------------------------------------------------------------------------------------------------------------------------------------------
[scalarkvsort.*random_128 vs. simdkvsort.*random_128]/uint64_t                +0.1078         +0.1102          1798          1992          1801          1999
[scalarkvsort.*random_128 vs. simdkvsort.*random_128]/int64_t                 +0.0765         +0.0833          1773          1909          1775          1923
[scalarkvsort.*random_128 vs. simdkvsort.*random_128]/double                  -0.2372         -0.2336          2155          1644          2157          1653
[scalarkvsort.*random_128 vs. simdkvsort.*random_128]/uint32_t                -0.3402         -0.3394          1689          1114          1691          1117
[scalarkvsort.*random_128 vs. simdkvsort.*random_128]/int32_t                 -0.3457         -0.3445          1705          1116          1706          1118
[scalarkvsort.*random_128 vs. simdkvsort.*random_128]/float                   -0.4488         -0.4485          2177          1200          2179          1201
OVERALL_GEOMEAN                                                               -0.2251         -0.2230             0             0             0             0

Benchmark                                                                      Time             CPU      Time Old      Time New       CPU Old       CPU New
-----------------------------------------------------------------------------------------------------------------------------------------------------------
[scalarkvsort.*random_1k vs. simdkvsort.*random_1k]/uint64_t                -0.2675         -0.2674         22952         16813         22963         16822
[scalarkvsort.*random_1k vs. simdkvsort.*random_1k]/int64_t                 -0.3494         -0.3493         23429         15243         23440         15253
[scalarkvsort.*random_1k vs. simdkvsort.*random_1k]/double                  -0.7164         -0.7162         44117         12513         44131         12523
[scalarkvsort.*random_1k vs. simdkvsort.*random_1k]/uint32_t                -0.6435         -0.6431         17177          6124         17184          6133
[scalarkvsort.*random_1k vs. simdkvsort.*random_1k]/int32_t                 -0.6460         -0.6458         17605          6231         17612          6238
[scalarkvsort.*random_1k vs. simdkvsort.*random_1k]/float                   -0.8430         -0.8428         44325          6960         44333          6968
OVERALL_GEOMEAN                                                             -0.6273         -0.6271             0             0             0             0

Benchmark                                                                      Time             CPU      Time Old      Time New       CPU Old       CPU New
-----------------------------------------------------------------------------------------------------------------------------------------------------------
[scalarkvsort.*random_1m vs. simdkvsort.*random_1m]/uint64_t                -0.8630         -0.8630     192462669      26358200     192446270      26357705
[scalarkvsort.*random_1m vs. simdkvsort.*random_1m]/int64_t                 -0.8728         -0.8728     192637350      24496545     192630780      24496142
[scalarkvsort.*random_1m vs. simdkvsort.*random_1m]/double                  -0.9052         -0.9052     217953698      20668767     217937330      20667636
[scalarkvsort.*random_1m vs. simdkvsort.*random_1m]/uint32_t                -0.9384         -0.9384     168347341      10363963     168335643      10363599
[scalarkvsort.*random_1m vs. simdkvsort.*random_1m]/int32_t                 -0.9362         -0.9362     170325919      10866820     170300695      10866190
[scalarkvsort.*random_1m vs. simdkvsort.*random_1m]/float                   -0.9425         -0.9425     196415775      11294391     196401933      11294226
OVERALL_GEOMEAN                                                             -0.9152         -0.9152             0             0             0             0

[r-devulap]

LGTM! thanks @sterrettm2!