perf(ext/web): optimize performance.measure()

[esroyo]

This PR optimizes the case when performance.measure() needs to find the startMark by name. It is a simple change on findMostRecent fn to avoiding copying and reversing the complete entries list.

Adds minor missing tests for:

• clearMarks(), general
• clearMeasures(), general
• measure(), case when the startMarks name exists more than once

Benchmarks

main

    CPU | AMD Ryzen 7 PRO 6850U with Radeon Graphics
Runtime | Deno 2.0.0-rc.4 (x86_64-unknown-linux-gnu)

benchmark              time/iter (avg)        iter/s      (min … max)           p75      p99     p995
---------------------- ----------------------------- --------------------- --------------------------
worst case measure()            2.1 ms         486.9 (  1.7 ms …   2.4 ms)   2.2 ms   2.4 ms   2.4 ms

this PR

    CPU | AMD Ryzen 7 PRO 6850U with Radeon Graphics
Runtime | Deno 2.0.0-rc.4 (x86_64-unknown-linux-gnu)

benchmark              time/iter (avg)        iter/s      (min … max)           p75      p99     p995
---------------------- ----------------------------- --------------------- --------------------------
worst case measure()          966.3 µs         1,035 (876.9 µs …   1.1 ms)   1.0 ms   1.1 ms   1.1 ms

Deno.bench("worst case measure()", (b) => {
  performance.mark('start');

  for (let i = 0; i < 1e5; i += 1) {
    performance.mark(crypto.randomUUID());
  }

  b.start();

  performance.measure('total', 'start');

  b.end();

  performance.clearMarks();
  performance.clearMeasures();
});

[CLAassistant]

All committers have signed the CLA.

[marvinhagemeister]

LGTM, that's a nice performance win!!

[nathanwhit]

LGTM, nice!