Praefect: replication jobs health ping

The replication process could take unpredictable amount of time. And to distinct between stale jobs (the process is finished, but the entry was not updated) and time-consuming jobs the health-ping was introduced. It updates timestamp of the event in the database so it can be used to identify if replication process is going on for this event.

Part of: #2873 (closed)

Follow up changes: !2322 (merged)

processBacklog function was refactored as it is hard to include a new things into it and hard to follow the complex logic it has in such a big method.