persisting machine annotations

[keighrim]

@marcverhagen suggested the other day that he wants to store output MMIFs from CLAMS pipeline initiated by an evaluation "invoker", perhaps somewhere on github, for later access. This would require

1. a schema to store pipeline configuration with the results
2. naming convention for results files
3. large enough storage to store all the results

This issue to discuss the design of such storage system and track progress of implementation.

[keighrim]

Just for future reference, while finalizing M3P2 period with our first ever round of app evaluation, we did upload machine annotation files (MMIF) by hand, under relevant xxx_eval directory, using preds@ prefix.