[ruff] Add rule to use human-readable names in ruff:ignore comments (RUF106)

[ntBre]

Summary

This is the second of our migration rules for moving from noqa comments to ruff:ignore comments
with human-readable names. RUF106 performs the second part of this transformation, converting
ruff:ignore comments that use rule codes to ruff:ignore comments with human-readable names.

The rule skips unknown codes, external and otherwise. Unlike RUF105 I don't think there's anything
sensible we could do with those here.

Test Plan

New mdtests and a CLI test for the combination of RUF105 and RUF106.

I did a local grep of our ecosystem projects and didn't find any ruff:ignore-style suppressions, so I don't expect any ecosystem hits for this.

[astral-sh-bot]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[MichaReiser]

This implementation is surprisingly simple.

I suggest changing the diagnostic range. Highlighting the entire ignore seems wrong to me:

• There's nothing wrong with the ignore itself, it's the codes
• There's nothing wrong with the reason
• There's nothing wrong with human-readable names within an ignore, but they're highlighted regardless.

[MichaReiser]

It would be nice if the rule only highlights the entire ignore if all codes are rule-codes. Otherwise, it would highlight consecutive runs of names.

However, I'd also be okay if we emit a separate diagnostic for each code. That's probably easiest. Given that this rule is primarily about the codes and not the ignore itself, I think that would be entirely reasoanble.

The reason why I think that'd be nicer is that it's not the entire ignore that's incorrect. You could also have a mixture of human-readable names and rule codes, in which case highlighting the entire ignore seems wrong to me

[ntBre]

I think I'll go for one diagnostic per code. That was actually my first instinct, but I went for this approach instead after seeing how RUF102 worked.

I do think the common case will be all codes, probably immediately after RUF105 runs on a noqa comment, but I still like separate diagnostics. I assume most noqa comments only have 1 code, in which case it's not that big of a difference either way.

Including the reason field was definitely an oversight in any case!

[MichaReiser]

I'd find it surprising that reason is highlighted here. There's nothing inherently wrong with the reason. I think we should just highlight the codes. We can try to group consecutive runs of codes, but I don't think we should highlight the entire ignore

[MichaReiser]

Can you add a test for an ignore that uses human readable names and codes? I don't think this is covered yet

[ntBre]

I think this is covered by:

ruff/crates/ruff_linter/resources/mdtest/ruff/rule-codes-in-suppression-comments.md

Lines 10 to 20 in 68ce969

	## `ruff:ignore`
	Rule names and external or unknown codes are preserved while Ruff rule codes are replaced:
	```py
	# snapshot: rule-codes-in-suppression-comments
	# ruff:ignore[F401, undefined-name, EXT001, UNKNOWN, F841]
	value = 1
	```
	```snapshot

if I understand your suggestion. Or do you mean just human-readable names and valid codes, without the external codes?

[MichaReiser]

yes, I mean one with valid human readable names and valid codes.

[ntBre]

Added a version without the external codes:

ruff/crates/ruff_linter/resources/mdtest/ruff/rule-codes-in-suppression-comments.md

Lines 53 to 60 in 20a6dd9

	Valid human-readable names are unaffected:
	```py
	# snapshot: rule-codes-in-suppression-comments
	# snapshot: rule-codes-in-suppression-comments
	# ruff:ignore[F401, undefined-name, F841]
	value = 1
	```