Split points classes

[PaulLeCam]

As discussed, this removes the single point model and instead make both the aggregation and allocation models use a shared interface.

The library now exports 3 reader classes:

• GenericReader to query documents
• SetReader which extends GenericReader with the ability to load deterministic (set account relation) documents
• PointsReader which wraps both GenericReader to query allocation documents and SetReader to load aggregation documents. Other than needing to use PointsReader.create() instead of new PointsReader(), the public APIs haven't changed.

It also exports 3 writer classes:

• ListWriter which extends GenericReader with the ability to create non-deterministic (list account relation) documents
• SetWriter which extends SetReader with the ability to update deterministic (set account relation) documents
• PointsWriter which wraps both ListWriter to interact with the allocation documents and SetWriter to interact with the aggregation documents. Other than needing to use PointsWriter.fromAuthenticated() instead of new PointsWriter(), the public APIs haven't changed.

[mzkrasner]

Looks good - thank you. I will need to update the server example to use the new aggregation interface.