[red-knot] fix incremental benchmark

[carljm]

Was it intentional that we rewrite foo.py with the BAR_CODE in the incremental benchmark?

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[MichaReiser]

I think the real error here is that we write /src/foo instead of bar. Because we then touch bar and not foo.

The idea is to invalidate a dependency and then re-run checking which should show how well we can cope with local changes.

[codspeed-hq]

CodSpeed Performance Report

Merging #12400 will degrade performances by 55.21%

_{Comparing cjm/fix-incremental-bench (5afe2a2) with main (f82bb67)}

Summary

❌ 1 (👁 1) regressions
✅ 32 untouched benchmarks

Benchmarks breakdown

	Benchmark	main	cjm/fix-incremental-bench	Change
👁	red_knot_check_file[incremental]	94.8 µs	211.7 µs	-55.21%

[carljm]

This change of course makes the incremental benchmark slower, since we now change the dependency and impact both files, requiring more validation of dependencies. It's still faster than the non-incremental benchmark though, which is good; the results of type inference on bar don't change due to the added comment and we can avoid redoing some work in foo.