GitLab

General strategy:
- perform all actions (sm_.Call)
- gather pending utterances
- mix in Expect* calls between all Call actions (incrementally from the top)
- keep the tests mostly simple (limit UI interactions) to prevent flakes
- do use the async SpeechMonitor api which should be less flakey
- deleted one test who's behavior is still in flux (empty panel menu behavior)

Some tests were still failing but due to external factors. Comments left.

Change-Id: I7689880dbeeac3b31fe22a8029fae5078d975f0a
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2119382
Commit-Queue: David Tseng <dtseng@chromium.org>
Reviewed-by: Akihiro Ota <akihiroota@chromium.org>
Cr-Commit-Position: refs/heads/master@{#753470}