clang-cl doesn't compile MSVC code using SIMD intrinsics

[dsharlet]

For example:

#include <immintrin.h>

__m128i f(__m128i a, __m128i b) {
  return _mm_mullo_epi32(a, b);  // SSE4.1 intrinsic
}

This builds using Microsoft's cl.

However, with clang-cl, we get: error: use of undeclared identifier '_mm_mullo_epi32'; did you mean '_mm_mullo_epi16'?

We can specify /arch:AVX to solve the problem. However, this means that the compiler is then free to use AVX instructions, which is not what we want if we want to only target SSE4.1. MSVC does not appear to have a /arch:SSE4.1 option: https://learn.microsoft.com/en-us/cpp/build/reference/arch-x64?view=msvc-170.

We can specify /clang:-msse4.1 to solve the problem, but this means that clang-cl is not a clean replacement of cl.

[zufuliu]

Replated and duplicate issues: #63492 and #56722.