[eradicate] Better detection of IntelliJ language injection comments (ERA001)

[InSyncWithFoo]

Summary

Enhances #14069.

Test Plan

A few tests are added to cover edge cases.

[InSyncWithFoo]

By the way, do we want that giant regex to be case-insensitive? type: ignore, pyright and SPDX-License-Identifier are a few non-examples.

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[MichaReiser]

Could you help me review this more efficiently by listing yoour changes so that I don't have to diff the Regex?

For case-sensitivity: I think we should use case-sensitive for the suppression comments that are case-sensitive and allow case-insensitivity for the ones that are not.

I also think we should gate this change behind preview mode unless all changes relax the rule's strictness.

[InSyncWithFoo]

Here's a comparison table:

	Before (all case insensitive)	After
pylint		No change
pyright		Case sensitive
noqa		No change
nosec		No change
region/endregion		Case sensitive
type: ignore		Case sensitive
fmt	fmt:\s*(on|off)	fmt:\s*(on|off|skip), case sensitive
isort		No change
mypy		Case sensitive
SPDX-License-Identifier:		Case sensitive
encoding		No change

As for language=:

• Before: (?i) language=[a-zA-Z](?: ?[-_.a-zA-Z0-9]+)+(?:\s+prefix=\S+)?(?:\s+suffix=\S+)?
• After: language=[-_.a-zA-Z0-9]+

[MichaReiser]

Thanks for the table. That made reviewing so much easier.

This looks good to me. The main open question is if the change should be gated behind preview. I'm leaning towards no, considering that there are no changes in the ecosystem check. @AlexWaygood are you okay with this?

[AlexWaygood]

I'm fine with this. It feels like a bugfix to me and, as you say, the ecosystem report indicates that it shouldn't be that disruptive.