clean up metadata approach

[jmccreight]

While #343 improved the link between pyPRMS metadata and pywatershed, the code is still very ad hoc and potentially difficult to to understand. The code could be refactored. Some other approaches could be considered as well. Among them, relying on pyPRMS to provide the metadata instead of parsing its XML files directly. This would require tests and stability on the pyPRMS side for the metadata class.