clang-cl should provide implementations of more MSVC ARM64 intrinsics

[rnk]

Some users have noticed that clang-cl does not provide some of the intrinsics listed in the MSVC ARM64 intrinsic list. In particular, the _Count* and _Copy* intrinsics seem worth implementing, like _CopyFloatFromInt32 and _CountLeadingZeros. See reference from the FP16 project here.

cc @amykhuang @pzhengqc

[compnerd]

@EugeneZelenko I assume that the windows label removal was accidental (if not please re-remove and leave a note as to why the MSVC behaviours are not really windows-y).

[EugeneZelenko]

@compnerd: Isn't clang-cl Windows-only thing?

[compnerd]

@EugeneZelenko yes, it is a windows thing, but can be used on other platforms. But that seems to indicate that it is supposed to be marked as windows (I was restoring the tag).

[davidbienvenu]

Another intrinsic that would be helpful is __prefetch.