[libc++] Try handling spurious cancellation in the mainline CI restarter

We've been having issues with cancelled CI jobs due to Docker-in-Docker
failures (spurious). I tried addressing that by adding a new flavor of
the restarter action, but that clearly hasn't worked out since we see
a lot of these spurious failures.

This PR is an attempt to fix this by changing the mainline CI restarter
used by libc++, not just the workflow job I added for testing. I have to
check this in to test it because this workflow uses the version that's
on the main branch.

.github/workflows/libcxx-restart-preempted-jobs.yaml

@@ -33,7 +33,7 @@ jobs:
         with:
           script: |
             const failure_regex = /Process completed with exit code 1./
-            const preemption_regex = /The runner has received a shutdown signal/
+            const preemption_regex = /(The runner has received a shutdown signal)|(The operation was canceled)/
 
             const wf_run = context.payload.workflow_run
             core.notice(`Running on "${wf_run.display_title}" by @${wf_run.actor.login} (event: ${wf_run.event})\nWorkflow run URL: ${wf_run.html_url}`)