Lithops: Do partial reprocessing when new databases are added

The Lithops annotation pipeline currently reprocesses all databases. This can cause significant repeated work. Copy across the logic that's in the Spark implementation:

• If the processing settings (the ds_config excluding the database field) has changed, always reprocess all DBs (IIRC this is also indicated by a reprocess flag with the annotate message)
• If databases have been removed, delete their corresponding job entries in postgres/elasticsearch but don't reannotate
• If databases have been added, annotate only the new databases

It would be a good idea to do #859 first, as this change will cause many centroids cache misses and #859 speeds up centroids cache regeneration.