resource scheduler: recovery on process restart

[sanderr]

The scheduler keeps some internal state. Some of it is directly derived from the DB state, but some of it is not. Examples include metadata related to event propagation (if any), ... This state should be somehow recoverable in case of failure.

• determine which state needs to be written to the DB for recovery
• design back up (may not be required, if everything is already in DB) + recovery approach + request review. Make sure to consider event recovery
• implement