This adds support for i32.shl, i32.shr_u and i32.shr_s.
These are the first instructions implemented which have constraints on
the registers they use (rcx in this case), so the implementation is a
bit more involved. It's still worth trying to emit good code here, as
shifts are quite common in our benchmarks.
This code will later have to be extended to use i32 immediates directly
instead of loading them into a register first. This will result in
smaller code and better performance.

R=titzer@chromium.org

Bug: v8:6600
Change-Id: I45b41ab062b58a9b2bc7e14a68663180307b900d
Reviewed-on: https://chromium-review.googlesource.com/859761
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Ben Titzer <titzer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50481}