libc: Add C23 stdbit.h with optional Kconfig and generic implementation

[arjav1528]

Fixes: #18311

Summary

Add optional C23 stdbit.h support so builds can provide <stdbit.h> when the toolchain does not. Follows the same pattern as stdarg.h / math.h: redirect at include/nuttx/lib/stdbit.h, copied to include/stdbit.h when enabled.

• Kconfig: ARCH_HAVE_STDBIT_H, ARCH_STDBIT_H (arch-specific), LIBC_STDBIT_GENERIC (generic on any arch).
• Header: Generic implementation using __builtin_clz/__builtin_ctz/__builtin_popcount for all C23 stdbit macros; archs may provide arch/<arch>/include/stdbit.h.
• Build: Unix.mk, Win.mk, CMake copy the header when either option is set.
• Libc: Optional libs/libc/stdbit/stdbit_verify.c (compile-time check when stdbit enabled).
• Docs: Documentation/legacy_README.md and Documentation/components/libs/libc/index.rst.

Impact

• Users: Optional; no change unless ARCH_STDBIT_H or LIBC_STDBIT_GENERIC is enabled. Code can then use C23 <stdbit.h> macros when the toolchain lacks them.
• Build: One header is copied to include/stdbit.h when stdbit support is enabled; build system (Unix.mk, Win.mk, CMake) is extended only for this copy.
• Compatibility: Improves portability for C23 stdbit usage on toolchains that don’t yet ship stdbit.h. No impact on existing configs that don’t enable it.

Testing

There is no freely available, upstream test suite for C23 stdbit.h that we could add to NuttX to validate this implementation. Verification is limited to:

• Enabling LIBC_STDBIT_GENERIC (or arch-specific ARCH_STDBIT_H where implemented) and confirming the build completes and stdbit_verify.c compiles (compile-time checks that the macros expand as expected).
• Manual inspection that the generic implementation matches C23 semantics and uses the intended builtins.

[michallenc]

That was fast, nice work! Please squash the commits ideally into two - one implementing stdbit and other updating the documentation. Also fix long lines in stdbit.h causing CI to fail.

We definitely need proper tests before merging this. These should go to app repository. I can write some, but won't get to it until next week probably.

[michallenc]

Seems the merge request now contains completely different commits unrelated to the title and previous changes.

[linguini1]

Please check that your PR follows the contributing guidelines. Also, a mistake might have happened during rebase/squashing that has caused your commits to disappear, as Michal pointed out.

[arjav1528]

Please check that your PR follows the contributing guidelines. Also, a mistake might have happened during rebase/squashing that has caused your commits to disappear, as Michal pointed out.

yep, that was a rebase issue, nvm I have reapplied the changes, you can check now

[linguini1]

Are there any test-suites freely available we could bring to NuttX to make sure this implementation works? I don't think we have any way to test it right now (not really a way around that though).

[michallenc]

I think we could write our own tests in apps directory. They should be pretty straightforward based on stdbit specification https://cppreference.com/w/c/header/stdbit.html. The only complicated part is to test it for every possible int size.

Not familiar enough with a method to test this implementation. Thank you for adding docs though!

Please add the Impact and Testing sections to your PR description (even if there's no way to test, just mention that).

[michallenc]

Just realized we should add common stdc_leading_zeros. It can be implemented as

#define stdc_leading_zeros(x) (stdc_leading_zeros_ull(x) - (unsigned int)(8 * (sizeof(0ULL) - sizeof(x)))

and similarly for other stdc defines.

[arjav1528]

Me too!, I have commited the change, kindly review the PR