Moderately straight-forward review. Refactoring and new ability to output Key Performance Metrics and their associated info as JSON files.

I would appreciate your thoughts on presenting the output. I currently have a separate JSON fild for each metric. Putting them together in one makes sense, but then would need a packaging solution, which was taking a step toward DM-2050. So I figured I would keep it simple and obvious and we can restructure things later however might be best useful.

Oh, right. J Matt Peterson [X] has wonderful reasons to not be reviewing tickets.

Tim Jenness Would you mind taking a look at this on his behalf?

I have some comments on the PR, many of them caused by confusion at lack of documentation for a parameter.

Regarding JSON output, keeping it simple is fine but I think we have to consider moving that generic serialization code to pipe_base.

Thanks Tim Jenness.
Merged to master.

Key changes in response to PR comments.
1. Documentation improved, particularly about the behavior of the plotBase, now outputPrefix, option.
2. Docstrings updated to better follow Python documentation standards.
3. JSON test now a real unittest.

Open Questions that arose in the review:
1. Should the Struct serialization to JSON code be adopted into pipe_base itself.
2. Should we encourage module-specific Error hierarchies, and if so, should those multiply inherit from both the module base error class and a relevant standard Python error class?