sort: retest l10n sorting changes with updated benchmarks

[ChrisDryden]

This PR was originally made by @sylvestre back in sept to measure the performance that adding localy sorting would have. When I was investigating why there was such a slowdown I found that the benchmarks were actually not working correctly because the locale would be set to the first locale in the tests instead of to what the env variable was set to.

Now that the benchmarks are updated, it shows that the performance decline for the c locale is negligible and that the only declines are when using the locale env variables. This it not even a "decline" per se because these sorts were not even working in the first place.

[codspeed-hq]

CodSpeed Performance Report

Merging #9995 will degrade performance by 94.31%

_{Comparing ChrisDryden:sort-perf-rebased (7531d74) with main (996e5e8)}

Summary

❌ 7 regressions
✅ 132 untouched
⏩ 37 skipped¹

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	BASE	HEAD	Efficiency
❌	sort_key_field[500000]	690.6 ms	1,090 ms	-36.64%
❌	sort_dictionary_order[500000]	1.1 s	1.1 s	-3.81%
❌	sort_german_de_locale	37.9 ms	665.9 ms	-94.31%
❌	sort_ascii_utf8_locale	20.1 ms	86 ms	-76.65%
❌	sort_reverse_utf8_locale	37.6 ms	609.2 ms	-93.82%
❌	sort_mixed_utf8_locale	37.4 ms	600.7 ms	-93.77%
❌	sort_unique_utf8_locale	37.7 ms	523.1 ms	-92.8%

Footnotes

1. 37 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[ChrisDryden]

@sylvestre how would you feel about merging your changes now that the benchmarks are corrected?

[sylvestre]

seems that https://github.com/uutils/coreutils/pull/9982/files#diff-e7bdf46e4a9e8a87200b91ae7a3eb7b406ac2d4388b70e108a7197ad93feaae9 is doing something similar for join
maybe we should add

pub fn should_use_locale_collation() -> bool {
    get_collating_locale().0 != DEFAULT_LOCALE
}

(or the equivalent) in uucore in a new pr

wdyt?

[ChrisDryden]

Yea that should definitely be in a shared place and have benchmarks to make sure that the more locale stuff we roll out does not impact performance. I'll make an example PR.

[sylvestre]

OK to land but please fix the jobs

[github-actions]

GNU testsuite comparison:

GNU test failed: tests/sort/sort-rand. tests/sort/sort-rand is passing on 'main'. Maybe you have to rebase?