Treat TypedDict (old-style) aliases as regular TypedDicts

[sterliakov]

Fixes #18692.

This PR makes mypy recognize old-style aliases to TypedDict types:

Alias = SomeTypedDict
ExplicitAlias: TypeAlias = SomeTypedDict

Still doesn't support generic no_args aliases:

from typing import Generic, TypedDict, TypeVar

_T = TypeVar("_T")

class TD(TypedDict, Generic[_T]):
    foo: _T

Alias = TD
# but works with
OtherAlias = TD[_T]

that's because no_args aliases are handled in code in several places and all of them expect such an alias to have Instance target.

[sterliakov]

This is rather important - stdlib stubs use TypedDict aliases (e.g. _SliceAttributes: typing_extensions.TypeAlias = _Attributes in ast.pyi where _Attributes is a generic TypedDict). We still don't support no_args aliases of generic TypedDict and NamedTuple types, they need to be parameterized with typevar explicitly. This may be a good future improvement, but now too many places require that no_args aliases have Instance as underlying type.

[github-actions]

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

hydra-zen (https://github.com/mit-ll-responsible-ai/hydra-zen)
- src/hydra_zen/structured_configs/_implementations.py:2950: error: Argument 2 to "make_dataclass" has incompatible type "**StrictDataclassOptions"; expected "str"  [arg-type]
- src/hydra_zen/structured_configs/_implementations.py:2950: error: Argument 2 to "make_dataclass" has incompatible type "**StrictDataclassOptions"; expected "tuple[type, ...]"  [arg-type]
- src/hydra_zen/structured_configs/_implementations.py:2950: error: Argument 2 to "make_dataclass" has incompatible type "**StrictDataclassOptions"; expected "dict[str, Any] | None"  [arg-type]
- src/hydra_zen/structured_configs/_implementations.py:2950: error: Argument 2 to "make_dataclass" has incompatible type "**StrictDataclassOptions"; expected "bool"  [arg-type]
- src/hydra_zen/structured_configs/_implementations.py:2950: error: Argument 2 to "make_dataclass" has incompatible type "**StrictDataclassOptions"; expected "str | None"  [arg-type]
- src/hydra_zen/structured_configs/_implementations.py:3310: error: Argument 2 to "make_dataclass" has incompatible type "**StrictDataclassOptions"; expected "str"  [arg-type]
- src/hydra_zen/structured_configs/_implementations.py:3310: error: Argument 2 to "make_dataclass" has incompatible type "**StrictDataclassOptions"; expected "tuple[type, ...]"  [arg-type]
- src/hydra_zen/structured_configs/_implementations.py:3310: error: Argument 2 to "make_dataclass" has incompatible type "**StrictDataclassOptions"; expected "dict[str, Any] | None"  [arg-type]
- src/hydra_zen/structured_configs/_implementations.py:3310: error: Argument 2 to "make_dataclass" has incompatible type "**StrictDataclassOptions"; expected "bool"  [arg-type]
- src/hydra_zen/structured_configs/_implementations.py:3310: error: Argument 2 to "make_dataclass" has incompatible type "**StrictDataclassOptions"; expected "str | None"  [arg-type]

[hauntsaninja]

Worth adding a test case for the PEP 695 alias?

[sterliakov]

@hauntsaninja We should already have a hard error whenever PEP695 type alias is used as a base class, is a test doing this specifically for TypedDict helpful?

[hauntsaninja]

Was a question, sounds like your answer is no! Thanks for the fix!