X64: replace explicit zero-extend with runtime assert

X64: replace explicit zero-extend with runtime assert

Since the input float64 is in the range of [0, 2^32-1], the upper 32 bits of output register
should be zero.

BUG=
R=bmeurer@chromium.org, titzer@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=25027

[Weiliang, 2014-10-10 08:48:28]

weiliang.lin@intel.com changed reviewers:
+ bmeurer@chromium.org, danno@chromium.org, dcarney@chromium.org,
titzer@chromium.org

[Weiliang, 2014-10-10 08:48:28]

PTAL

[titzer, 2014-10-15 12:10:53]

lgtm

[Benedikt Meurer, 2014-10-17 09:48:48]

I think this is NOT LGTM. We should really get the 32 vs 64 int/uint stuff right
in the graph.

[Benedikt Meurer, 2014-10-31 09:18:15]

Ok, so Ben convinced me that this is the way to go. So, LGTM. Will land.

[Benedikt Meurer, 2014-10-31 09:20:00]

Committed patchset #1 (id:1) manually as 25027 (presubmit successful).