Redesign the Cell morphology data storage model

[mwatts15]

The Cell morphology is currently written such that it expects the morphology to be extracted from NeuroML XML files and rewritten as triples in a fairly naive fashion. It's needed to reevaluate the need for cell morphology in owmeta to determine what approach would be best for storing cell morphology.

The approach taken must allow to efficiently recover a prototypical cell morphology for any given cell from owmeta in neuroml format if sufficient data exists to produce such a cell.

Some possible approaches:

• Modify the earlier tactic in such a way that the data can be efficiently queried from a triple store and turned into neuroml Python objects directly (including storing the XML directly in the graph or storing a compressed form of it)
• Store references to NeuroML XML describing C. elegans cells. For example, create a URI scheme which references existing NeuroML XML files in CElegansNeuroML repo describing morphology of typical named cells. Annotate the URI reference as needed:
  • to point to accessible files (if the URI scheme doesn't point to a location where the file can be accessed),
  • to relate aggregate facts about the morphology, like maximum physical extents (the "bounding box"),
  • to give example values for free variables in the morphology.
• A step beyond the first approach, storing the data/models on which the generated NeuroML in CElegansNeuroML is based.

The last of these approaches is most consistent with current development in c302, and likely needs done anyway, but the first or second also have value for recovering morphologies for use of cell morphologies without regenerating them.