All configuration must be available via OPTIONS in Spark SQL's CREATE TEMPORARY TABLE

[Spiro Michaylov]

Much of NSMC's configuration uses SparkContext properties, but these don't work over, say, a JDBC connector when the context is not easily accessible, and in anyc ase they make it hard to use a diverse set of MongoDB instances and even collections.

Spark SQL integration needs to be extended to allow all of this configuration to be passed into the OPTIONS clause of CREATE TEMPORARY TABLE, etc, one a per-collection basis.