Consistently handle self-resolving alerts

The following discussion from !52374 (merged) should be addressed:

• @syasonik started a discussion:

  Should "resolved" alerts which have not been received before be created & resolved? Or just skipped, as it is now for Prometheus alerts?

  Non-prometheus alerts could have a resolving payload, but the alert will be created without the status change or ends_at time being applied. Should these be resolved? Or skipped completely?