mechababs aggregate

Produce a strawman for yarik to review:
- what data gets collected
- some cases, stuff thats not committed, (babs jobs_status.csv things)
- slurm jobs
- resource usage of slurm jobs (must be collected before slurm drops them from history)
- how that data gets rendered

Concepts to consider:
- cheap render-only mode ie from cache
- query/collect mode ie needs to discover the reality first
- look into git statusd
- where cache lives git-annex uses a special branch to record the "state of the world"
- fscacher might be helpful

Goal: Dashboard "What are we doing, and how done is it"
- list all mechababs campagins
- for a campaign, what studies are involved
- for each study
- what derivatives are complete
- do we need to more info if partially complete?
- record babs specific info, like "has these results branches"?
- these jobs are running/done/pending?
- full list of jobs with resource usage?
- interaction with "ephemeral studies", ie what if the user removes the study from the cluster?

Start by reviewing the existing mechababs data sources, including babs jobs_status.csv and Slurm job history, then investigate git statusd, git-annex's special branch, and fscacher. Produce a strawman for yarik covering collection, caching, rendering, campaign/study/job views, resource usage, and ephemeral studies; completion means the proposed design is ready for review.

Written by the indexing model from the issue text.