[ty] Add micro-benchmark for #711

[sharkdp]

Summary

Add a benchmark for the problematic case in astral-sh/ty#711, which will potentially be solved in #18955

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[codspeed-hq]

CodSpeed Instrumentation Performance Report

Merging #18979 will not alter performance

_{Comparing david/benchmark-711 (390c45b) with main (e5e3d99)}

Summary

✅ 37 untouched benchmarks
🆕 2 new benchmarks
⁉️ 1 (👁 1) dropped benchmarks

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
🆕	ty_micro[complex_constrained_attributes_1]	N/A	56.8 ms	N/A
🆕	ty_micro[complex_constrained_attributes_2]	N/A	6.3 s	N/A
👁	ty_micro[many_attribute_assignments]	56.7 ms	N/A	N/A

[AlexWaygood]

Oof... I just ran the benchmarks locally to see if a change I was experimenting with had a performance impact, and this new benchmark takes over a minute to run (much longer than any other ty benchmark):

Benchmarking ty_micro[complex_constrained_attributes_2]: Warming up for 3.0000 s
Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 85.2s, or reduce sample count to 10.
ty_micro[complex_constrained_attributes_2]
                        time:   [808.20 ms 808.88 ms 809.60 ms]
Found 1 outliers among 100 measurements (1.00%)
  1 (1.00%) high mild

The command I ran locally was cargo bench -p ruff_benchmark --bench=ty. For comparison, the next-slowest benchmark was project/anyio, for which the report was:

Benchmarking project/anyio: Warming up for 3.0000 s
Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 14.5s, or reduce sample count to 30.
project/anyio           time:   [131.96 ms 132.08 ms 132.23 ms]
                        change: [−0.9888% −0.7683% −0.5652%] (p = 0.00 < 0.05)
                        Change within noise threshold.
Found 4 outliers among 100 measurements (4.00%)
  1 (1.00%) low mild
  1 (1.00%) high mild
  2 (2.00%) high severe

Can we make this slightly less diabolical for now, so that it's possible to run benchmarks locally in a reasonable time?

[sharkdp]

Oof... I just ran the benchmarks locally to see if a change I was experimenting with had a performance impact, and this new benchmark takes over a minute to run (much longer than any other ty benchmark):

Oh, sorry! Will fix that later. In the meantime, you can use a filter argument or reduce the amount of samples (--sample-size 10), I think. The true fix will be #18955

[sharkdp]

Will fix that later

#19001