Suppress misleading Foo[...] suggestion for annotations with keyword args

[Chessing234]

Summary

• Fixes Mypy shoould not suggest Foo[...] if annotation contains Foo(arg=...) etc. #16506
• When a call expression with keyword arguments is used in a type annotation (e.g., Foo(arg=val)), mypy previously suggested use Foo[...] instead of Foo(...). Since Foo[arg=val] is a SyntaxError in Python, this suggestion is misleading.
• This PR suppresses the bracket syntax suggestion when the call node contains keyword arguments. The "Invalid type comment or annotation" error is still emitted, just without the unhelpful note.

Changes

• mypy/fastparse.py: Added a check for e.keywords in visit_Call so the Suggestion: use Foo[...] instead of Foo(...) note is only shown when there are no keyword arguments.
• test-data/unit/check-fastparse.test: Added test case testFasterParseTypeErrorCustomWithKeyword verifying no suggestion note is emitted for Foo(arg=int).
• test-data/unit/check-errorcodes.test: Added test case testErrorCodeSyntaxError4 verifying the error code behavior with keyword arguments.

Test plan

• testFasterParseTypeErrorCustom_no_native_parse still passes (existing behavior preserved for positional args)
• testFasterParseTypeErrorCustomWithKeyword_no_native_parse passes (new test for keyword args)
• testErrorCodeSyntaxError4_no_native_parse passes (new error code test)
• All testErrorCodeSyntaxError* tests pass

🤖 Generated with Claude Code

[github-actions]

Diff from mypy_primer, showing the effect of this PR on open source code:

ibis (https://github.com/ibis-project/ibis)
- ibis/backends/datafusion/udfs.py:22: note: Suggestion: use dt.Timestamp[...] instead of dt.Timestamp(...)
- ibis/backends/datafusion/udfs.py:30: note: Suggestion: use dt.Timestamp[...] instead of dt.Timestamp(...)
- ibis/backends/datafusion/udfs.py:42: note: Suggestion: use dt.Timestamp[...] instead of dt.Timestamp(...)
- ibis/backends/datafusion/udfs.py:50: note: Suggestion: use dt.Timestamp[...] instead of dt.Timestamp(...)
- ibis/backends/datafusion/udfs.py:111: note: Suggestion: use dt.Timestamp[...] instead of dt.Timestamp(...)

[wyattscarpenter]

Your two tests seems redundant with each other 🤷‍♂️

[wyattscarpenter]

Repeating my suggestion from the main issue discussion, since it's also relevant here:

Assuming we can't/won't make calling the function in the annotation just work, if the problem mypy is trying to help you with is that people write Foo(arg=x) when they mean Foo[x], then isn't an additional suggestion called for? Like, keep the note: Suggestion: use Foo[...] instead of Foo(...), but add an additional note: Cannot use named arguments as type arguments. This could even fire orthogonally, as another note on invalid-type, whenever named arguments are used. According to 87f1588 this should be a pretty simple feature to implement. If we can make both of those not appear, and be replaced with note: Can't call a function in a type annotation if mypy determines the call is actually to a function, even better.