Node execution with predecessor data

[reelmatt]

This builds on the read/write of Node data to files @diegostruk1 introduced in PR #35.

The key changes are in pyworkflow/workflow.py and are:

• Moves (most) file-handling to the Workflow object, using the FileStorage API. Node data is written using store_node_data() and read using retrieve_node_data()
• Adds an execute(node_id) method to the Workflow that orchestrates Node execution. This works by
  • Retrieving a node_to_execute specified by node_id
  • Retrieving data from any preceding Nodes
  • Passing that data to the Node's execute() function
  • Writing that output to a file and storing the filename in the Node's data attribute

Includes some basic error-checking/exception-handling, but not a lot. You can test execution using sample_execution.zip.

1. Move join1.csv and join2.csv to /tmp
2. Load execution_workflow.json into Postman
3. Manually execute Nodes 1-4. You should see the file names output to screen, and the "Retrieve data" test should return the data written to /tmp.

Edit: latest commits add more error-checking/handling. Namely, endpoints should now return a 400/500 code with message for missing data files/improper execution for most cases.

[diegostruk]

Looks good!