[match-case] Fix narrowing of class pattern with union-argument.

[randolf-scholz]

Fixes #19468

Based on earlier PR #19473

• refactored the conditional_types function.
  • return UninhabitedType(), default when no ranges are given. This corresponds to isinstance(x, ()), with an empty tuple, which always returns False at runtime.
  • when the type is a union, factorize the yes type isinstance(A|B, C) -> yes = A_yes | B_yes (note: this does not apply to the no-type generally, see [match-case] Fix narrowing of class pattern with union-argument. #19473 (comment)
• modified [match-case] Fix narrowing of class pattern with union-argument. #19473 to change the proposed type, rather than directly returning. This is essential to maintain the behavior of the unit test testIsinstanceWithOverlappingPromotionTypes
• Allow TypeChecker.get_isinstance_type to return empty list (fixes isinstance(x, ()) behavior).

Modified tests

• testIsInstanceWithEmtpy2ndArg now correctly infers unreachable for isinstance(x, ()).
• testNarrowingUnionMixins now predicts the same results as pyright playground, https://mypy-play.net/?mypy=latest&python=3.12&gist=734c95f128705cfb0dc4ce24f1ad8eec

New Tests

• testMatchNarrowDownUnionUsingClassPattern (https://mypy-play.net/?mypy=1.17.0&python=3.12&gist=e9ec514f49903022bd32a82ae1774abd)

[randolf-scholz]

This test is actually redundant, it works just as well if we comment it out since make_simplified_union([]) returns UninhabitedType().

[github-actions]

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

spark (https://github.com/apache/spark)
- python/pyspark/ml/connect/readwrite.py:100: error: Redundant cast to "JavaEstimator[Any]"  [redundant-cast]
- python/pyspark/ml/connect/readwrite.py:103: error: Redundant cast to "JavaEvaluator"  [redundant-cast]

prefect (https://github.com/PrefectHQ/prefect)
- src/prefect/tasks.py:559: error: Incompatible types in assignment (expression has type "Union[int, float, list[float], None]", variable has type "list[float]")  [assignment]
+ src/prefect/tasks.py:559: error: Incompatible types in assignment (expression has type "Union[float, int, list[float], None]", variable has type "list[float]")  [assignment]

starlette (https://github.com/encode/starlette)
+ starlette/middleware/errors.py:176: error: Incompatible types in assignment (expression has type "Any | None", variable has type "Response")  [assignment]
+ starlette/middleware/errors.py:176: error: Argument 1 has incompatible type "Request"; expected "WebSocket"  [arg-type]
+ starlette/_exception_handler.py:59: error: Argument 1 has incompatible type "Request | WebSocket"; expected "WebSocket"  [arg-type]
- starlette/_exception_handler.py:61: error: Argument 1 to "run_in_threadpool" has incompatible type "Callable[[Request, Exception], Response | Awaitable[Response]] | Callable[[WebSocket, Exception], Awaitable[None]]"; expected "Callable[[Request | WebSocket, Exception], Any]"  [arg-type]
+ starlette/_exception_handler.py:61: error: Argument 1 to "run_in_threadpool" has incompatible type "Callable[[Request, Exception], Response | Awaitable[Response]] | Callable[[WebSocket, Exception], Awaitable[None]]"; expected "Callable[[Request | WebSocket, Exception], Any | None]"  [arg-type]

hydra-zen (https://github.com/mit-ll-responsible-ai/hydra-zen)
- src/hydra_zen/wrapper/_implementations.py:933: error: Incompatible return value type (got "type[object]", expected "DataClass_ | type[DataClass_] | ListConfig | DictConfig")  [return-value]

urllib3 (https://github.com/urllib3/urllib3)
+ src/urllib3/_collections.py:359: error: Redundant cast to "Iterable[tuple[str, str]]"  [redundant-cast]

websockets (https://github.com/aaugustin/websockets)
+ src/websockets/legacy/protocol.py:641: error: Redundant cast to "Iterable[str | bytes | bytearray | memoryview[int]]"  [redundant-cast]

[hauntsaninja]

Thanks!