Range based locks for fdbdr

The current locks for fdbdr lock the entire database. This makes it difficult to have multiple different ranges from different clusters replicated to the same destination.

First we need to separate locks into locks that prevent reading a range and locks that prevent writing a range. When DR is still copying data from the primary we want to prevent both reads and writes. Once the DR database is a full replica of the primary, we only want to prevent writes to the range.

Proposed design:
* Store the locks in the \xff keyspace, and send changes to this list of locks to clients with read version requests.
* Clients check individual reads and writes against the list.

Start by tracing fdbdr's current database-wide locking and how clients receive read version requests. Review the proposed lock list in the \xff keyspace, separating read-blocking and write-blocking ranges, then follow how clients check reads and writes. Done means replication blocks both operations during copying and only writes after the database is a full replica, while supporting different ranges from multiple clusters.

Written by the indexing model from the issue text.