Basic kuduRDD and default source

I'm creating this jira to capture the change we made to the original Spark integration plan, and to have a reference jira for 0.7.0.

What we have now is:

1. a kuduRDD that uses newAPIHadoopRDD with our input format.
2. a default source with a base relation that basically just wraps the kuduRDD, meaning you can run SparkSQL and pick your columns but that's it.
3. a unit test infrastructure in scala

I'm assigning it to Dan since he's the one who took it to the finish line.