Temporary intentional performance regression for animations

[Rune Lillesveen]

Hi, sorry for x-posting.

There's an assert in an optimization for animations which checks if
the optimization is sound. Yesterday I landed a CL[1] turning that
assert into a release-assert temporarily on trunk to try to catch
remaining issue. Turning that assert into a release assert means that
the optimization will effectively be canceled out.

1. If you see performance regressions for animations, this is a likely
cause. cc ru...@opera.com in the bug reports.
2. If you see the RELEASE_ASSERT(*m_baseLayoutStyle == *layoutStyle);
happening, cc ru...@opera.com in the bug reports.

[1] https://codereview.chromium.org/1001323003/

--
Rune Lillesveen