Trust Tokens: LOG(ERROR), not FATAL, on bad command-line key commitments

Right now we LOG(FATAL) when a user provides bad Trust Tokens key
commitments through the command line (for local debugging). This was
intended to crash Chrome entirely when it's provided a syntactically
invalid command line flag during local testing, with the thought that
the user would want to know immediately that there was a problem with
their testing setup. Sadly, LOG(FATAL)ling from the network service
unfortunately just crashes the network service, not Chrome as a whole;
this network service crash is not directly user-visible and can lead to
confusing behavior like no pages loading.

This change alters the log line to a LOG(ERROR) so that the failure
remains consistently visible in the command line console without leading
to hard-to-diagnose behavior like pages failing to load.

Bug: None

Change-Id: I70a40b1057140f17754f4c8b479f64256d1bc54c
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2285466
Commit-Queue: Steven Valdez <svaldez@chromium.org>
Reviewed-by: Steven Valdez <svaldez@chromium.org>
Auto-Submit: David Van Cleve <davidvc@chromium.org>
Cr-Commit-Position: refs/heads/master@{#786280}

services/network/trust_tokens/trust_token_key_commitments.cc

@@ -36,10 +36,10 @@ ParseCommitmentsFromCommandLine() {
           raw_commitments)) {
     return std::move(*parsed);
   } else {
-    // Crash loudly here because the user presumably only provides key
+    // Complain loudly here because the user presumably only provides key
     // commitments through the command line out of a desire to _use_ the key
     // commitments.
-    LOG(FATAL)
+    LOG(ERROR)
         << "Couldn't parse Trust Tokens key commitments from the command line: "
         << raw_commitments;
   }