Batch query support

[dpratt]

Self-explanitory.

The trick here is to figure out how to support a batch of queries that may have disparate keyspaces associated with them. There are a couple routes we could take - if a batch request comes in, we could partition it into sub-batches with homogeneous keyspaces, and aggregate the results as they come back. Alternatively, we could just punt (as it appears that the java driver does) and use the keyspace definition from the first query in the batch.