For the sake of BSON-NumPy and other libraries that can consume raw BSON, add an API that retrieves raw batches as undelimited streams of BSON documents. (A sequence of RawBSONDocument instances is too slow in some circumstances, see PYTHON-1284.)

Proposed new methods are find_raw_batches and aggregate_raw_batches. These methods will return new cursor types which are iterables of batches. Each batch is a raw document stream: some number of undelimited raw BSON bytes all catenated together.