Concentrated header for internal bit utilities

[frederick-vs-ja]

Also reduces inclusion dependency on <limits>.

Currently __isa_available and _Stl_isa_available_meow are made unconditionally declared, but it might be better to avoid declaring them under some circumstances.

Fixes #3692. Fixes #832.

MSVC-internal changes are needed for the new header <__msvc_bit_utils.hpp>.

[StephanTLavavej]

No change requested: Nice catch!! We mistakenly lacked _INLINE_VAR here:

STL/stl/inc/limits

Lines 1204 to 1205 in c8d1efb

	template <class _Ty>
	constexpr bool _Is_standard_unsigned_integer =

[StephanTLavavej]

I'm mirroring this to the MSVC-internal repo - please notify me if any further changes are pushed.

[StephanTLavavej]

I pushed a conflict-free merge with main, then I had to push a commit to guard the lzcnt etc. intrinsic usage because /clr:pure (and CUDA/Intel) doesn't understand them. We didn't need it before when this was in <bit> which is C++20, but now this code is being promoted to C++14 mode.

[StephanTLavavej]

Note to other reviewers: I kept this structure for consistency with our other intrinsics guards, even though this one could arguably be De Morganed.

[StephanTLavavej]

Thanks for this major header restructuring! 🛠️ 🎉 😸

[StephanTLavavej]

I observe that <bitset> and <numeric> are still indirectly dragging in <limits> which we probably should have checked in the first place, but this is still a reasonable restructuring.

[frederick-vs-ja]

I observe that <bitset> and <numeric> are still indirectly dragging in <limits> which we probably should have checked in the first place, but this is still a reasonable restructuring.

<numeric> includes <xutility>, which is made dragging <limits> in since C++23 by #3678. I'm trying to find a way to avoid the indirect inclusion. In many cases we only need numeric_limits<I>::max(), perhaps we should invite a variable template for it.