Data loader for analysis architecture

We need to build a data loader for the analysis architecture. This should accept as input a pipeline or collection and write the corresponding data table with one row for every data asset that was run. The table will have at minimum the columns:

- input asset name, as "asset_name" -- this will be a shared column used to join with other tables
- pipeline
- collection
- S3 results folder location

And then any processing input and output parameters as additional columns.

I'm not quite sure how this works for pipeline versioning, etc, but that's the basic idea. The loader should live in a subfolder of https://github.com/AllenNeuralDynamics/zombie/tree/dev/src/zombie/data, probably zombie/data/analysis_architecture/.