Add retry on hadoop client calls

[raronson]

If the cluster is under a lot of load, client calls which directly interact with it can fail in weird ways. I have seen LeaseExpiredException, ConnectException: Connection refused, and others which all point to a cluster under load not able to respond to requests.

Scoobi should retry these client calls a number of times so the app doesn't die on the first exception from a hadoop call.

Examples:

• Counters
• HDFS interaction
• Other?