Support RBS signatures in check-shims command

[sambostock]

Motivation

The check-shims command detects duplicate definitions between shim files and generated RBIs. It correctly ignores duplicates when the shim adds a sig block (since the shim is adding type information, not just duplicating). However, it doesn't recognize RBS comment syntax (#:) as a signature, so shims using RBS syntax are incorrectly flagged as duplicates.

With Sorbet's RBS support, users can now write type signatures using RBS comments instead of sig blocks. The check-shims command should treat these equivalently.

Implementation

Modified has_duplicated_methods_and_attrs? in RBIFilesHelper to check for RBI::RBSComment instances in a node's comments array, in addition to checking the sigs array. The logic mirrors the existing signature handling:

• A shim with an RBS comment is not considered a duplicate of an untyped method (it's adding type info)
• A shim with a different RBS signature is not a duplicate (it's refining the type)
• A shim with an identical RBS signature is flagged as a duplicate

Added a small helper method extract_rbs_comments to extract RBS comments from a node.

Tests

Added 3 new tests mirroring the existing sig block tests:

• ignores duplicates that have an RBS signature
• ignores duplicates that have a different RBS signature
• detects duplicates that have the same RBS signature

[amomchilov]

Suggested change

	return true if shim_rbs_comments.any? { |rbs| other_rbs_comments.include?(rbs) }
	return true if shim_rbs_comments.intersect?(other_rbs_comments)

[sambostock]

Unfortunately intersect? won't work here because RBI::RBSComment implements == but not eql?/hash:

c1 = RBI::RBSComment.new('() -> void')
c2 = RBI::RBSComment.new('() -> void')

c1 == c2           # => true
[c1].include?(c2)  # => true  (uses ==)
[c1].intersect?([c2])  # => false (uses eql?/hash)

I've added a comment to explain this.

[egiurleo]

@sambostock Not blocking this PR, but could we make RBI::RBSComment implement .eql? or is there something preventing that?

[egiurleo]

@sambostock I suspect you'll need to rebase to get CI fixes. I'll investigate further if that doesn't fix things.