Handle type[TypeVar] (fixes #14755)

[sterliakov]

Fixes #14755.

This case is already covered by check-classes.test::testTypeUsingTypeCClassMethodFromTypeVarUnionBound, now we just remove from it errors that shouldn't be reported.

[github-actions]

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

steam.py (https://github.com/Gobot1234/steam.py)
- steam/app.py:129: note: Maybe you forgot to use "await"?
- steam/chat.py:401: error: No overload variant of "__new__" of "type" matches argument type "Type[ChatMessageT]"  [call-overload]
- steam/chat.py:401: note: Possible overload variants:
- steam/chat.py:401: note:     def __new__(cls: Type[type], object, /) -> type
- steam/chat.py:401: note:     def [_typeshed.Self] __new__(cls: Type[_typeshed.Self], str, Tuple[type, ...], Dict[str, Any], /, **kwds: Any) -> _typeshed.Self

urllib3 (https://github.com/urllib3/urllib3)
+ test/__init__.py:97: error: Module has no attribute "PROTOCOL_TLS"  [attr-defined]
+ test/__init__.py:97: note: Error code "attr-defined" not covered by "type: ignore" comment

[sterliakov]

The changes seem legit:

• steap/app.py: I don't fully understand, there is no error for that line, why it used to be a note there?
• steam/chat.py: there was a case of A.__new__(A), where A is some class - correct, was a bug, because A was defined by TypeVar with union upper bound (here)
• urllib3: the error was simply not reported before (see added else branch: before this PR it was possible to fall through context.has_no_attr without emitting error message) - here.

[hauntsaninja]

Thanks, this looks great!

re: steam/app.py, there is an error for that line, it's just that the error existed both before and after this change so it doesn't show up in the diff. Why the note goes away isn't exactly clear to me, I will take a look later

[hauntsaninja]

Thank you!