Add missing unit tests for # noqa: A-like cases

[InSyncWithFoo]

Summary

Follow-up to #16659.

This change adds tests for these three cases, which are (also) not covered by existing tests:

• # noqa: A (lone incomplete code)
• # noqa: A123, B (complete codes, last one incomplete)
• # noqa: A123B (squashed codes, last one incomplete)

The third case is mentioned in the specification, but neither of the other two are:

If an item does not consist entirely of codes (e.g. F401abc), parsing aborts with an error and the user is warned of an InvalidCodeSuffix [...]

Test Plan

Unit tests.

[codspeed-hq]

CodSpeed Performance Report

Merging #16677 will degrade performances by 10.49%

_{Comparing InSyncWithFoo:noqa (f644fbb) with micha/ruff-0.10 (e9bfdfd)}

Summary

❌ 1 regressions
✅ 31 untouched benchmarks

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

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
❌	red_knot_check_file[incremental]	4.9 ms	5.5 ms	-10.49%

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[dylwil3]

Thanks!

The third case is mentioned in #16483 (comment), but neither of the other two are

These cases are covered by this part of the spec:

We expect a list of rule codes, separated by commas or whitespace. After splitting first on commas, and then on whitespace, we attempt to parse each item in turn. If parsing a nonempty item returns no rule codes, we stop. If no codes were found at all we warn the user with MissingCodes.

e.g. A is not a valid code, so we stop. There were no preceding codes, so we return MissingCodes, etc.