Skip unknown tables during incremental snapshots

When running incremental snapshots, unknown tables will cause the snapshotting to be cancelled. Once Debezium is restarted, it will go to retry the snapshot and get cancelled on the exact same table. In the case of a typo or something similar, snapshotting will be held up until the operator catches the issue.

Instead, the table could be skipped over and a jmx metric logged to let the operator know that something is wrong.  

Similarly, tables that don't have a primary key should be skipped over instead of causing Debezium to restart.