GitLab

I've been investigating why the realbox has such worse ToPaint metrics
than the omnibox. I discovered that the realbox is ignoring all
autocomplete results other than the final "done()" one, which won't come
in until we get results from the Suggest server, often 60ms away at the
fastest.

This CL makes the realbox paint incomplete results, just like the
omnibox does. For the user, it's going to make the realbox feel much
more responsive, like the omnibox, but also introduces the same
asynchronous result-changing-after-the-fact issue that the omnibox has.

I think it's best to just keep the behavior the same between the UIs.

It also improves the ToPaint time on my machine from ~70ms to ~15ms,
which is quite competitive with the actual omnibox.

Some history:
In the original implementation of the realbox, there was a one-time-use
callback, which is probably why the original implementer waited until
the autocomplete machinery was done().

This limitation was removed by this CL:
https://chromium-review.googlesource.com/c/chromium/src/+/1875030

This CL takes advantage of that new ability by reporting the incomplete
states as well. (i.e. history provider has returned but search suggest
has not gotten the remote results yet).

Bug: 1046561
Change-Id: I8f7fc5760bfefb0e9aa0d3642cde0e30185e71be
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2036259Reviewed-by: Moe Ahmadi <mahmadi@chromium.org>
Commit-Queue: Tommy Li <tommycli@chromium.org>
Cr-Commit-Position: refs/heads/master@{#739184}