GitLab

This is a simpler approach to tackle (workaround) the blocked network
thread issue when compared to CL 2436278.

This CL simply creates a base::Watchdog that arms before closing the
peer connection, and disarms after the peer connection is closed. If
thread join takes longer than expected then it will crash the host so
that the daemon process can respawn it.

Bug: 1130090
Change-Id: Ib2400e17b848ed07df902a02b344efbf5e15c588
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2446840Reviewed-by: Lambros Lambrou <lambroslambrou@chromium.org>
Commit-Queue: Yuwei Huang <yuweih@chromium.org>
Cr-Commit-Position: refs/heads/master@{#813937}