As part of the openstack decomission, we need to rework our test flow. Instead of relying on a stable runner, watching over long-running jobs in beaker, allow the runner to suffer unscheduled reboots by splitting the test submission and external test watchdog. This can be done by saving the "recipe" somewhere, and resuming the watchdog. Jira: [CKI-7097](https://issues.redhat.com/browse/CKI-7097)