GitLab

This found a bug in saturated_cast that was covered by ClampToInt tests
so it fixes that.

If you have a floating point value of MAX_INT, then comparing an
integer against it for equality with promote it to a float and
it will compare true. However if you cast the float to an int, its
actually outside the bounds of integer, so you can end up with a
negative int as a result. Added unittests to check this for
saturated_cast.

Review URL: https://codereview.chromium.org/1164063005

Cr-Commit-Position: refs/heads/master@{#333092}