GitLab

Previously, we would fade out the fakebox and fade in the real omnibox during
the scroll transition, which created discrepancies. With this change, the
toolbar omnibox does the full transition. The search box on the NTP is still
kept, both to get the starting point of the transition, and for the tablet
toolbar, where the omnibox stays at the top, so there are two boxes.

Because the two boxes now have to match, their dimensions and padding are
slightly updated. The height is 56dp, and the padding on the sides is 12dp.

Also, in the focused state the omnibox now has 1dp bleed at the sides,
which moves the rounded corners out of view.

The remaining changes are not user-visible:
* Opacity of the fakebox is now driven by the toolbar. This lets us centralize
  the logic (the toolbar can decide its own opacity and the opacity of the
  fakebox) and use different behavior on tablets, where the fakebox will simply
  fade out as before.
* Rename variables to (hopefully) make them more consistent / self-explanatory.
* |mLocationBarBackgroundBounds| (formerly |mUrlViewportBounds|) more accurately
  reflects the visible omnibox bounds in the absence of the NTP / when the NTP
  is scrolled so that the omnibox is at the top. To do that, we no longer
  include the Y translation we apply to position it where the fakebox is (it's
  instead applied to the |mLocationBarBackgroundNtpOffset|), and we always use
  an expansion value of 1 if the current tab is an NTP.
* Remove the |inset_textbox| drawable -- it was the same as |textbox|, but with
  a margin added at the top and bottom (but *not* left or right). Since we
  calculate the omnibox bounds manually anyway, this allows us to account for
  the omnibox padding in the same way in either dimension.
* Remove the |isInTabSwitcherMode| parameter from some methods in favor of using
  the member variable.

BUG=605054,625108,618955,616728,612520

Review-Url: https://codereview.chromium.org/2134663002
Cr-Commit-Position: refs/heads/master@{#408885}