Database backed merkle tree implementation

[HarryR]

Instead of storing the whole tree in-memory, the Python implementation should be able to store the nodes on-disk using a fast key-value store.

This will allow for tree sizes to exceed available memory.

The storage backend should be pluggable.

Potential database backends:

• https://rocksdb.org/
• sqlite
• leveldb
  • python wrapper: https://github.com/wbolster/plyvel
• http://sophia.systems/
  • python wrapper: https://github.com/coleifer/sophy

The database interface should be able to commit batches of inserts/updates to improve performance