Better provenance tracking

[JoanneBogart]

Currently a single config file, including provenance, is written when one or more output data files are created.
To be sure of what we're getting, provenance information needs to be kept on a per-data-file basis. It should include at a minimum

• code identification. At least one of release version, git hash
• run-time options and their values
• inputs
• output schema version
  Ideally the information should be stored as metadata in the binary output.