borealis: fix for metrics when borealis shuts down

Currently startup metrics are recorded every time the VM is
shutdown using the functionality in the context manager. This
is incorrect, we should only be recording startup metrics when
a startup attempt has completed. The change makes it so that
shutdowns that cancel a startup attempt (noted by startup tasks
existing) emit a metric, but other shutdowns do not.

Bug: none
Change-Id: Iecb5c91e143f3a0550f9bc4bec7283f7c1bebfbd
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2577079Reviewed-by: Nic Hollingum <hollingum@google.com>
Commit-Queue: Daniel Ng <danielng@google.com>
Cr-Commit-Position: refs/heads/master@{#834965}

chrome/browser/chromeos/borealis/borealis_context_manager_impl.cc

@@ -118,6 +118,7 @@ void BorealisContextManagerImpl::Complete(BorealisStartupResult result,
                                           std::string error_or_empty) {
   startup_result_ = result;
   startup_error_ = error_or_empty;
+  if (result != BorealisStartupResult::kCancelled || !task_queue_.empty())
     RecordBorealisStartupResultHistogram(result);
 
   while (!callback_queue_.empty()) {