Add neon_cnt and neon_addv8 for 64-bit ARM for parity with MSVC

[llvmbot]

Bugzilla Link	51488
Version	12.0
OS	Windows NT
Reporter	LLVM Bugzilla Contributor
CC	@DougGregor,@zygoloid

Extended Description

Hello.
I started to work on popcount optimization on ARM64: microsoft/STL#2127
and I found linking errors with clang-cl.
Could you provide neon_cnt and neon_addv8? (Also neon_cntq, neon_addvq8 for future optimizing bitset: microsoft/STL#2129 (comment))

__n64  neon_cnt(__n64);
__n128 neon_cntq(__n128);
__n8  neon_addv8(__n64);
__n8  neon_addvq8(__n128);

Or do you already provide similar intrinsics with different names?
I am slightly worried that they are not _Ugly.
So maybe I should ask Microsoft as well, maybe they should rename it if you already have it with suitable names...

I used Windows 10 x86-64, Visual Studio 2022 Preview 3, Clang 12 from Visual Studio.

[StephanTLavavej]

MSVC's STL is now using different intrinsics, so we've closed this issue.