reconsider using ZooKeeper for partitioned job database

[soxofaan]

The current implementation of the "partitioned jobs" feature (large area processing, cross-backend processing, ...) depends on ZooKeeper for internal housekeeping.
At the time of the original implementation, this was the most obvious choice as we were also using it for the job registry. This job regustry solution did not scale very well so we migrated to an Elastic Search based solution.
Partitioned job execution is not yet commonly used, so it's not highly urgent, but at some point we should consider eliminating this dependency and replace it, e.g. with ES or SQLite