Port IsValidOp improvements from JTS

[pramsey]

Port of locationtech/jts#743 and locationtech/jts#748.

[dbaston]

I'm seeing a pretty big performance regression on my Australia test-case:

main:

➜  cmake-build-release git:(main) bin/perf_unary ~/data/australia.txt valid
Reading geometries from /home/dan/data/australia.txt
Read 1 geometries.
625,265 usec

This PR:

➜  cmake-build-release git:(c7d40c15f) bin/perf_unary ~/data/australia.txt valid
Reading geometries from /home/dan/data/australia.txt
Read 1 geometries.
21,245,292 usec

[pramsey]

Ew. Supposed to be faster... hopefully I have some terrible hotspot that'll show up in profile. Also, I'm sure @dr-jts will check your case in JTS and see if it's an algorithmic woopsie.

[dbaston]

➜  jts git:(master) ✗ bin/jtsop.sh -time -a ~/data/australia.txt isValid
	
Operations: 1  Total Time: 78.8 s

Maybe the new algorithm removes some of the indexing that GEOS was using? e.g. #256

[dbaston]

Here's my data. Nothing special, just an extract from GADM.
australia.txt.gz

[dr-jts]

Maybe the new algorithm removes some of the indexing that GEOS was using? e.g. #256

That could well be it. The Nested Shells check in the new IsValidOp code is using a nested loop, so is O(n^2). Easy to fix. I will look at that ASAP.

[dr-jts]

JTS has added IndexedNestedPolygonTester to improve performance of that check, See JTS-755.

[pramsey]

Merged at e151d23