- Eliminates I8x16SPLAT macro. This generates:

  push eax
  mov  eax, #imm32
  movd xmmDest, eax
  pxor xmmScratch, xmmScratch
  pshufb xmmDest, xmmScratch
  pop eax

- Uses smaller sequence for I8x16Shl and I8x16ShrU that doesn't
  require a temp register.
- Uses punpckbw, psraw, packsswb sequence for I8x16ShrS.

Bug: v8:6020
Change-Id: I1dc4ecaba43b150a3412e9083fe930868d9e5d5d
Reviewed-on: https://chromium-review.googlesource.com/1121045Reviewed-by: Jing Bao <jing.bao@intel.com>
Commit-Queue: Bill Budge <bbudge@chromium.org>
Cr-Commit-Position: refs/heads/master@{#54300}