Account for selector specificity when merging extend_unsafe_fixes and override extend_safe_fixes

[lukaspiatkowski]

Summary

Prior to this change extend_unsafe_fixes took precedence over extend_safe_fixes selectors, so any conflicts were resolved in favour of extend_unsafe_fixes. Thanks to that ruff were conservatively assuming that if configs conlict the fix corresponding to selected rule will be treated as unsafe.

After this change we take into account Specificity of the selectors. For conflicts between selectors of the same Specificity we will treat the corresponding fixes as unsafe. But if the conflicting selectors are of different specificity the more specific one will win.

Test Plan

Tests were added for the FixSafetyTable struct. The check_extend_unsafe_fixes_conflict_with_extend_safe_fixes_by_specificity integration test was added to test conflicting rules of different specificity.

Fixes #8404

[zanieb]

Thank you for contributing! I have some questions about simplifying the implementation but otherwise this looks great.

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[lukaspiatkowski]

I was going to ask if there is a chance this will make it to the next release, but I see the release happened yesterday :) is this good to go as is @zanieb?

[zanieb]

Sorry for the delay this one got lost!

[lukaspiatkowski]

LGTM