Develop specification for post-processing APIs

[trevorhardy]

Right now there are several distinct capabilities around post-processing that are similar but largely independently developed. We need to evaluate these capabilities and design a post-processing API that incorporates as much of these existing capabilities but refactors to be a coherent, designed interface. This task is to evaluate those capabilities and design the API; the implementation will be done in a later task.

Considerations in the design:

• We want to be able to create an abstraction between reading in the data from various formats (HDF5, databases) and working on read-in data. This requires creating a data schema definition that all the readers create and all the post-processing works on. I propose this is a Pandas Dataframe with a particular schema. We will define unique schemas for each data source.
• We don't need to worry about making GridLAB-D metrics-collector work with JSON outputs. These can produce VERY large JSONs that aren't tractable to use. HDF5 is better in all ways and we need to get cross-platform support for it in GridLAB-D.
• Though we'll likely start with working on GridLAB-D data, we need to look at the other data sources including market data and bulk power system simulation data. There is existing data collection for all of these that we should consider. Or maybe for now we just worry about GridLAB-D.