This ticket involves 4 changes:

1. Keep a database size field on the connection object
2. On checkpoint completion write out the checkpoint size in the checkpoint metadata struct.
3. On checkpoint pickup read in the checkpoint size 
4. Maintain a running total of b-tree sizes throughout checkpoint, for b-trees that are then checkpointed subsequently apply the delta of the size difference back to the database level size.

All of these changes can be found on the PoC branch, including a python test that tests various database size level scenarios.