On ia32, the implementation of 64-bit shifts first moved {src} into
{dst}, then {amount} into {ecx}. This fails if {dst} overlaps {amount},
because {amount} would be overwritten before being used. Just changing
the order to these two moves would also not be correct, since {src} can
contain {ecx}.
Thus, implement this via a general parallel register move, which
resolves cycles automatically.

R=titzer@chromium.org

Bug: v8:7589, v8:6600
Change-Id: I2556b9aa66a89a067372b7713dbbb3d71d2f923f
Reviewed-on: https://chromium-review.googlesource.com/981134
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Ben Titzer <titzer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#52327}