Consider use of `pyarrow.lib.RecordBatchReader` for internal streamable data handling

[d33bs]

DuckDB enables the use of pyarrow.lib.RecordBatchReader returns which could help us speed up or optimize certain internal processing for CytoTable. We could consider this as time goes on to help with performance improvements.

References #369