Checker runs aren't cancelled if a cluster is removed during the run

Steps to reproduce:

1. Add a cluster
2. Wait for a checker run (either automatic or manual)
3. While it's running, remove the cluster from cbmultimanager

Even though the cluster is removed, some checker results will persist in the store, which will mean that e.g. Alertmanager tries to create alerts for a non-existent cluster and fails (potentially failing the entire Alertmanager run).

(Relatedly, it looks like clusterUUID in checkerResults doesn't have a foreign key to the checkers table, which would have prevented this.)