While we already do have logic to re-run flaky tests, the FAILED
variable currently does not get reset to "0". As a result, successful
reruns will still cause the test to be registered as failed.

Fix this by resetting the variable accordingly.