Amend annotation for asyncio.Protocol.__slots__

[mikeshardmind]

This fixes an assignablity issue in subclasses that was caused by using too strict an annotation for the purpose of __slots__ as a special case for mypy.

We can keep an annotation here for mypy if really needed, but it needs to work with subclass assignability.

ideally, long term all data model names are understood as part of the specification, but this needs to be permissive enough to allow subclasses to add members.

See error in pyright as a result of adding a special case for mypy: microsoft/pyright#11058

[AlexWaygood]

Thanks! It looks like there are a few other instances that we should probably fix similarly -- could you maybe do those too?

~/dev/typeshed (main)⚡ [1] % git grep '__slots__: tuple\[()\]'
stdlib/asyncio/protocols.pyi:    __slots__: tuple[()] = ()
stdlib/asyncio/protocols.pyi:    __slots__: tuple[()] = ()
stubs/protobuf/google/protobuf/internal/well_known_types.pyi:    __slots__: tuple[()] = ()
stubs/protobuf/google/protobuf/internal/well_known_types.pyi:    __slots__: tuple[()] = ()
stubs/protobuf/google/protobuf/message.pyi:    __slots__: tuple[()] = ()

[mikeshardmind]

sure, is it better in this case to put them all in this PR, or will it work better for history if these are seperate?

[AlexWaygood]

I think we added them all in one PR, so it probably makes sense to fix them all in one PR. I also don't mind if you'd rather split them up, though!

[mikeshardmind]

All in one seems simpler then. That was a much smarter grep than what I used, I just checked for the known comment, figuring it would have been replicated to explain anywhere that was used for that reason.

[AlexWaygood]

thanks!

[github-actions]

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉