Solve issues with deferral; related minor corrections

[LordGiacomoS]

Fixes crash in mypy (python/mypy#21588) that caused deferral of this branch's PR (python/typeshed#13740) on the main typeshed, because while the crash could be solved in mypy, it seems more prudent to use this typeshed-based solution. Solutions that would modify mypy's code would introduce a redundant version check and still have the crash if an up-to-date version of typeshed isn't used. The exact changes that fix the crash are specifically the changes to stdlib/types.pyi in 76cf69a).

Other minor corrections:

• get_origin overload changes (changes to stdlib/typing.pyi & stdlib/typing_extensions.pyi in 40d9f63) were done to solve conflicting overloads issue occuring during ./tests/mypy_test.py stdlib when Union changes are in effect.
  • tested to work with mypy despite changes, even in case of ParamSpecArgs or ParamSpecKwargs instances being provided.
• changes to builtins.type.__or__ & builtins.type.__ror__ (changes to stdlib/builtins.pyi in 40d9f63) were made to fix an issue with how mypy displays reveal_type calls on pep 604 unions (e.g. reveal_type(str | int) -> types.UnionType | def () -> int).
• remaining changes (changes to stdlib/types.pyi in 40d9f63; all changes in 96f1bf1) are minor amendments to fix incredibly minor details and make the linter happy.

Remaining Concerns:

• Mypy still raises inaccurate concerns about Unions when running test cases (currently working on this one)
• Pyright still raises inaccurate concerns about Unions when running test cases (next on my to-do list)
• stubtest errors introduced by Viicos's change and still not fixed
  • typing.Union.__name__ & typing.Union.__qualname__: Given that it is valid when running mypy over code accessing these specific class properties, I don't believe this check is actually necessary, as they seemingly already exist implicitly. I'm not confident enough in this belief to touch things though.
  • unused allowlist entry notes:
    • typing.Union: It doesn't seem like classes themselves are able to be allowlist entries, so maybe this should be something like typing.Union.*?
    • others: I don't think any of these are actually necessary (or possible) to implement, unless checking for a negative, as the methods don't exist independently of typing.Union

Stubtests: https://github.com/LordGiacomoS/typeshed/actions/runs/29355015820 (runs a different branch to allow me to use modified version of mypy to fix more minor errors that aren't in scope of crash specifically)
Tests: https://github.com/LordGiacomoS/typeshed/actions/runs/28957923025

Stubtest errors excerpt

Sourced from: https://github.com/LordGiacomoS/typeshed/actions/runs/28900665825/job/85736371504

error: typing.Union.__name__ is not present in stub
Stub: in file /home/runner/work/typeshed/typeshed/stdlib/typing.pyi
MISSING
Runtime:
'Union'

error: typing.Union.__qualname__ is not present in stub
Stub: in file /home/runner/work/typeshed/typeshed/stdlib/typing.pyi
MISSING
Runtime:
'Union'

note: unused allowlist entry typing.Union
note: unused allowlist entry types.UnionType.__class_getitem__
note: unused allowlist entry types.UnionType.__mro_entries__
note: unused allowlist entry types.UnionType.__name__
note: unused allowlist entry types.UnionType.__qualname__
Found 7 errors (checked 700 modules)

[LordGiacomoS]

Latest commit (#2eccc22 ) fixes issue found when running test cases on python 3.14 & 3.15, across all operating systems

Issue in question

Source: https://github.com/LordGiacomoS/typeshed/actions/runs/29354539513/job/87161643214#step:4

mypy --platform {linux, darwin, win32} --python-version {3.14, 3.15} on the standard library test cases: FAILURE

stdlib/@tests/test_cases/check_types.py:82: error: Expression is of type
"<typing special form>", not "Never | Any"  [assert-type]
        assert_type(union_type | Literal[1], types.UnionType | Any)
        ^

[LordGiacomoS]

The changes to builtins.type.__or__ & builtins.type.__ror__ (changes to stdlib/builtins.pyi in 40d9f63) currently break mypy's handling of unions and that fact, combined with how messy this branch is (due to tacking on a number of other commits as a brute force approach to solving a specific problem) and the lack of visibility/feedback here, leads me to believe that closing this PR, cleaning up the changes, and opening one on the mainline typeshed repo is the best approach to solve these issues.