- Add I16x8 Splat, ExtractLane, ReplaceLane, shift ops, Some BinOps and compare ops
 - Add pshufhw, pshuflw in the assembler, disassembler
 - Fix incorrect modrm for pextrw, this bug disregards the register allocated and always makes pextrw use rax.
 - Fix pextrw disasm to take the 0 - 7 bits of the immediate instead of 0 - 3.
 - Pextrw, pinsrw are in the assembler use 128 bit encodings, pextrw, pinsrw in the disassembler use legacy encodings, fix inconsistencies causing weird code gen when --print-code is used.

Review-Url: https://codereview.chromium.org/2767983002
Cr-Commit-Position: refs/heads/master@{#44664}