Vector algorithms cleanups

[AlexGuteniev]

Easy cleanups:

• More constants. (Ideally, make everything constant, except loop counter and loop-carried dependency).
• _M_IX86 / _M_X64 to _WIN64. That's the last occurrence after Enable ARM64EC vector algorithms #6084.
• _mm_cvtsi64x_si128 -> _mm_cvtsi64_si128. Standardize spelling.

Standardizing for _mm_cvtsi64_si128 was preferred over the other way around because:

• We already consistently prefer _mm_cvtsi128_si64 over _mm_cvtsi128_si64x
• Clang does not understand the si64x form of either
• The form without x is more common (searched on GitHub and elsewhere)
• The x character doesn't seem to convey any meaning. Seems that it is supposed to mean that general-purpose 64-bit integer is used, as opposed to __m64 MMX vector, this can be seen in _epi64x ending intrinsics, which have _epi64 counterparts, But this is no longer ever necessary, as no point in using MMX.

[StephanTLavavej]

I'm mirroring this to the MSVC-internal repo. Please notify me if any further changes are pushed, otherwise no action is required.

[StephanTLavavej]

Thank you for your ongoing dedication to code quality and consistency! 🪄 😺 💯