gh-157212: Fix _Py_ThreadId() on Windows ARM64 with MinGW

[christianaurichzm]

Fixes #157212

_Py_ThreadId() currently uses __getReg(18) for the MinGW ARM64 branch:

#elif defined(__MINGW32__) && defined(_M_ARM64)
    tid = __getReg(18);

That intrinsic is not available for MinGW ARM64. For a *-windows-gnu target, Clang's intrin.h defers to mingw-w64's header, which declares __getReg only for Itanium. GCC does not provide a corresponding builtin either.

The branch is reached by AArch64 MinGW toolchains because mingw-w64 defines _M_ARM64 when __aarch64__ is defined.

Falling through to the generic __aarch64__ implementation is not a valid alternative. The Windows ARM64 ABI reserves x18 as the platform register pointing to the Thread Environment Block (TEB). In the system tested in gh-157212, x18 also matched NtCurrentTeb() for each tested thread, while tpidr_el0 was observed as zero.

This change therefore reads the TEB address directly from x18 for MinGW
ARM64:

__asm__ ("mov %0, x18" : "=r" (tid));

The MSVC branch is unchanged.

Verification

I reproduced the original compile failure with zig cc/Clang targeting
aarch64-windows-gnu:

error: call to undeclared function '__getReg'; ISO C99 and later do not
support implicit function declarations [-Wimplicit-function-declaration]

With this change, the same target compiles successfully and at -O2 emits the expected register read:

mov     x0, x18
ret

The change is limited to the MinGW ARM64 branch.

AI tools were used to help organize and edit the wording of this PR description.

[kumaraditya303]

I think _Py_get_machine_stack_pointer has the same issue on ARM64 Windows MinGW.

[corona10]

lgtm!

https://godbolt.org/z/sj97f6TEr

I've double checked with godbolt code gen :)

[miss-islington-app]

Thanks @christianaurichzm for the PR, and @corona10 for merging it 🌮🎉.. I'm working now to backport this PR to: 3.13, 3.14, 3.15.
🐍🍒⛏🤖

[miss-islington-app]

Sorry, @christianaurichzm and @corona10, I could not cleanly backport this to 3.14 due to a conflict.
Please backport using cherry_picker on command line.

cherry_picker c1af94ae348da826ac530c8d5097fa887e19579c 3.14

[miss-islington-app]

Sorry, @christianaurichzm and @corona10, I could not cleanly backport this to 3.13 due to a conflict.
Please backport using cherry_picker on command line.

cherry_picker c1af94ae348da826ac530c8d5097fa887e19579c 3.13

[bedevere-app]

GH-158164 is a backport of this pull request to the 3.15 branch.