GitLab

Testing with Windows Narrator scan mode on Wikipedia showed we were
spending an inordinate amount of time comparing text positions due to
the cost of creating ancestor positions. This CL avoids that cost by
comparing the order of the first uncommon ancestors for each node, when
it's possible to do so.

Testing on https://en.wikipedia.org/wiki/List_of_highest-grossing_films
with Narrator scan mode heading movement showed a reduction of about 87%
CPU usage on the browser main thread with this CL.

Bug: 1029867
Change-Id: I7efae50517a561319dad81668c49ec5f6ae900b5
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1999083
Commit-Queue: Kevin Babbitt <kbabbitt@microsoft.com>
Reviewed-by: Dominic Mazzoni <dmazzoni@chromium.org>
Reviewed-by: Jacques Newman <janewman@microsoft.com>
Reviewed-by: Adam Ettenberger <adettenb@microsoft.com>
Reviewed-by: Ethan Jimenez <ethavar@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#732522}