GitLab

In the new world where we are no longer uniquifying keywords, and
instead tolerating multiple engines with the same keyword (we pretty
much have to), it's more convenient for us to use an std::multimap to
map keywords to engines rather than use std::map.

This is to unblock a few ideas I have:

 - Currently, to detect conflicts between engines, we traverse through
   the whole list of TemplateURLs during Add() and Update(). During
   startup and Sync, when we are adding a bunch of these engines, that
   makes the whole process O(n^2). I'd like to use this multimap to
   remove those loops (in a future CL).

 - Although it SEEMS like we may no longer need to detect conflicts in
   the new world of tolerating conflicts, that's not true. We still
   want to remove replaceable conflicting engines, and so we still want
   to detect them.

 - I also hope the multimap will enable me to unify and simplify the
   currently complicated logic that does the removal of replaceable
   conflicting engines.

 - Finally, I think the multimap won't be expensive to retrieve the
   "best" engine for any given keyword, since the typical number of
   values we linearly go through per keyword is going to be one to
   three... definitely single digits. Multimaps don't sort on value.

These things are going to be in followup CLs that are enabled by the
multimap.

See also:
https://chromium-review.googlesource.com/c/chromium/src/+/2519775

Bug: 1022775
Change-Id: I631ea6c3cf442736273e23fcda15f72095dca28a
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2530801
Commit-Queue: Tommy Li <tommycli@chromium.org>
Reviewed-by: Peter Kasting <pkasting@chromium.org>
Cr-Commit-Position: refs/heads/master@{#826545}