It seems like there weren't any dEQP tests for constant folding of
nonsquare matrices component-wise multiplication. There were a couple
bugs in our implementation which could lead to undefined behaviour.

Fixes the code and cleans up a few style issues.

Also includes a regression test.

Bug: chromium:912505
Bug: chromium:912508
Change-Id: I7fb85d1404a32950fa9fe4c3bbba9edc9f38ddd1
Reviewed-on: https://chromium-review.googlesource.com/c/1387065
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
(cherry picked from commit 2b35654d)
Reviewed-on: https://chromium-review.googlesource.com/c/1390363Reviewed-by: Jamie Madill <jmadill@chromium.org>