Add batching/chunking to aggregate.

[stefancoe]

Currently, aggregating large datasets, especially with longer distances, can use up a lot of memory. Add the ability to break up data into chunks and run them through aggregate sequentially.