use increasing retry intervals for hmshandler and metastoreclient retries

We should have smaller retry interval for the initial attempts, and longer intervals for subsequent attempts. This will avoid hitting the database or metastore server too hard when things go bad.
Exponential backoff or its variant can be used.