use owned ast and tokens in bench

[MichaReiser]

Summary

Use owned values for the linter benchmark as preparation for the parser refactor

The perf regressions come from the fact that the tokens and ast are now dropped inside the benchmark. I think this is good, because that is closer to what we see in production.

[codspeed-hq]

CodSpeed Performance Report

Merging #11598 will degrade performances by 15.62%

_{Comparing use-owned-ast-and-tokens-in-bench (9517a4d) with main (7659114)}

Summary

❌ 5 (👁 5) regressions
✅ 25 untouched benchmarks

Benchmarks breakdown

	Benchmark	main	use-owned-ast-and-tokens-in-bench	Change
👁	linter/all-rules[large/dataset.py]	16.2 ms	16.9 ms	-4.32%
👁	linter/default-rules[large/dataset.py]	3.6 ms	4.3 ms	-15.62%
👁	linter/default-rules[numpy/ctypeslib.py]	904.6 µs	1,011.7 µs	-10.58%
👁	linter/default-rules[pydantic/types.py]	1.9 ms	2.1 ms	-12.29%
👁	linter/default-rules[unicode/pypinyin.py]	350.7 µs	405.2 µs	-13.47%

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[dhruvmanila]

Thanks!