In our codegen, we are mixing SSE and AVX. This can potentially cause
transition delays. Ideally we should stick to one. We add some new AVX
versions of movss and movsd, then use the macro-assembler methods to
generate AVX instructions if supported.

Bug: v8:11190
Change-Id: Iff7c0fb892cea85731f880ac2895480621b3092f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2554257Reviewed-by: Bill Budge <bbudge@chromium.org>
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#71349}