Add peephole optimization to replace x86 SIMD instructions with smaller equivalents

[saucecontrol]

As pointed out in #112728 (comment), there are some SIMD instructions that have equivalents with smaller encoding.

For example, pupcklpd can be replaced with movlhps, which is 1 byte smaller, under the following conditions:

1. both operands are from register
2. both operands are 16-byte
3. no EVEX options (e.g. embedded masking) are used.

Similarly, vpermilps can be replaced with the smaller vpshufd if mixing float and integer domain instructions is not a concern.

llvm has logic that identifies some of these replacements here

cc @tannergooding

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-meta
See info in area-owners.md if you want to be subscribed.