Create *private* RIA store

[mih]

To be put on bulk1 eventually. This will be a superset of the one from #38 and additionally contain the pristine raw data and other unreleased or unreleasable components.

Most importantly it will provide a clean separation of the "dirty" workspace on juseless, and the "managed" parts of the dataset.

Datasets to move:

• [x] gumpdata
• [x] 7T_ad
• [x] pandora
• [x] anatomy
• [x] 3T_av_et
• [x] 3T_visloc
• [x] aggregate
• [x] aligned (untracked qa/ dir, converted to proper dataset)
• [x] annotations
• [x] comp_stim_features
• [ ] emoreg (no result were saved, massive untracked stuff, IT WAS NO ANNEX DATASET !??!)
• [x] fg_eyegaze (deleted, was a test for phase2)
• [x] fg_eyegaze_raw (untracked rescue/)
• [x] freesurfer (converted to proper dataset)
• [x] media (untracked files)
• [ ] anondata (phase1)
• [x] phase2
• [x] retmap (converted to proper dataset)
• [ ] sandbox (not a dataset, connected to the timeline annotation tool)
• [x] structural
• [x] tnt (some untracked qa data in git-ignored dirs)
• [x] visloc (untracked condor_logs)
• [x] studyforrest_avmovie_shen_aggregate (deleted, contained in aggregate)
• [ ] speech-annotation

TODO:

Copy all previously existing datasets (again) with git-annex copy --all to make sure that alternative branches and previous versions remain intact!!!

Done for all datasets that are checked above, except: https://github.com/datalad/datalad/issues/5613 happening for tnt and visloc

[mih]

The store is done, but the issue linked above prevents some data from entering