Clean up Workflow methods; update naming for Workflow/edit endpoint

[reelmatt]

This cleans up pyworkflow/workflow.py by removing unused methods, and condensing others. Namely, there was from_session, from_file, and from_request methods that all read in a Workflow in mostly the same way. Now, there is one from_json method that can load a Workflow from a file (via the 'open' endpoint) and from session (via the middleware).

Workflow attributes have been simplified to:

• name: Name of the Workflow (now defaults to model.options.id passed in from the front-end). An 'edit_workflow' endpoint has also been added to make this a more user-friendly name.
• root_dir: Added in #47, but Django references have been removed from pyworkflow to increase modularity of the package. Workflow.path() now handles creating valid path names for reading/saving files.
• graph: the computational graph
• flow_vars: the global flow variables, stored in a separate Graph() object.

These changes should not affect the existing functionality, at least for tests that contain all the required info. One area I wasn't able to test that might be affected is cases where there is missing info. The from_json and to_session_dict methods for reading/writing the Workflow now assume valid data is provided and will raise a WorkflowException if it's not there. Previously, in some situations it would assign None to the graph if a graph was not present, but that's not really the behavior we want (I think). Something to check/look out for.

[reelmatt]

Good note re: UUID/naming. I moved the comment I was writing here to #51 as a way of making it a TODO.

[reelmatt]

Changes since PR:

• Made path() instance method
• Moved root_dir logic to a new WorkflowUtils class
• Left generate_file_name, retrieve_node_data, and store_node_data as-is, but these may be due for changes later on as we figure more of the passing of data between nodes.

I think it makes sense to merge this now, and if we want to revisit these points later on, it should be easy enough to do.