887 select_dtypes stubs fixing

[mutricyl]

• Closes In DataFrame.select_dtypes type values like np.number or np.datetime64 should be allowed #887
• Tests added: Please use assert_type() to assert the type of any return value

I tried include with overloads limitations stated in select_dtypes documentation:

ValueError

• If both of include and exclude are empty
• If include and exclude have overlapping elements
• If any kind of string dtype is passed in.

I managed to cover somehow the first tick covering when neither of include or exclude is provided but I am not sure that it is worth the introduced complexity in stubs.

[Dr-Irv]

Overall, this looks good. I have ideas for you to try to catch the case of passing str or str as the dtype, as well as the 2 empty list case, so see if that works.

I don't think that overlapping lists can be detected with static typing.

[Dr-Irv]

I think if you added the following here as the first 2 overloads, you could catch the case of the string arguments not being allowed:

@overload
def select_dtypes(self, include: StrDtypeArg, exclude: AstypeArgExtList | None = ...) -> Never: ...
@overload
def select_dtypes(self, include: AstypeArgExtList | None = ..., exclude: StrDtypeArg) -> Never: ...

And if you added this, I think that you'd catch the case of 2 empty lists:

@overload
def select_dtypes(self, include: list[Never], exclude: list[Never]) -> Never: ...

You import Never from typing_extensions

There's a possibility that either mypy and/or pyright will complain about overlapping overloads, but you can add #type: ignore and # pyright: ignore as needed in the stubs to ignore that complaint.

[mutricyl]

That is implemented. Never is actually stronger than I expected since pylance read the stubs and visually show that the program will stop there.

[mutricyl]

@Dr-Irv your comments are implemented, you can review this PR one more time.
Regards

[Dr-Irv]

thanks @mutricyl . Nice use of assert_never()