HBase SPI for DDL

For replication, it would be useful to have an SPI that users can implement for executing HBase DDL operations. The base implementation would be whatever is done today, operating on a single HBase instance. Another implementation could call an external service that operates on multiple HBase clusters and sets up replication between the clusters.

Full design at https://wiki.cask.co/display/CE/HBase+DDL+SPI.