This is failing fairly frequently because the GC we trigger in the first tab
for the gcJSInitiatedDuring test is finishing sooner than we expect.
I think what's happening is that this GC is being finished by a timer
triggering GC, or possibly some other activity doing so. I added logic to retry
the test up to three times if the first GC finishes sooner than expected. This
makes the test pass on try now. I added the same logic to the
gcJSInitiatedBefore which looks like it could fail in the same way.
I removed the GC zeal parts because this test is only run on opt builds, which
don't have GC zeal enable by default.
Differential Revision: https://phabricator.services.mozilla.com/D247100