Reland: crash upon bad message in network service process in test build

This is a reland of http://crrev.com/c/6203517, which was reverted due
to a test hitting the bad message. That issue has been fixed by
http://crrev.com/c/6243909 and we should be able to reland the change to
crash network service process upon bad mojo message in non official
builds.

As bad mojo message including the one triggered by kIncorrectLock should
not happen outside of tests or compromised renderer process, switch to
crash in non official builds when it happens so that potential code bugs
will be more visible.

Unit tests that expect bad message override bad message handler and
should not hit this crash.

Add kIgnoreBadMessageForTesting switch for browser tests that expect to
trigger bad message to ignore the reported bad message so that we don't
crash network service process when test is expected to trigger bad
message, and updated RenderFrameHostImplBrowserTest to use the switch so
that its URLLoaderFactoryNotTrusted test does not fail due to expected
triggering of bad message.

Note that we don't want to crash network service process upon bad
message from compromised renderer process. Therefore we don't use CHECK
or NOTREACHED() which now crashes. There is also still bugs like
http://crbug.com/40066149 that could cause the bad message. Therefore,
we strictly limit crashing to non official builds.

Bug: 40823083
Change-Id: I95c3cadadf1d7b2179099a4e9eac64d3e76bd162
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/6270418
Commit-Queue: Liang Zhao <lzhao@microsoft.com>
Reviewed-by: Kenichi Ishibashi <bashi@chromium.org>
Reviewed-by: Takashi Toyoshima <toyoshim@chromium.org>
Reviewed-by: Greg Thompson <grt@chromium.org>
Reviewed-by: Rakina Zata Amni <rakina@chromium.org>
Cr-Commit-Position: refs/heads/main@{#1421518}