Meta-issue: upstream babs dependencies (part of M2)

Tracks the upstream `PennLINC/babs` dependencies that gate **M2-Correct-Publishable**. These are filed and worked upstream; this epic surfaces them in the milestone plan. Source of truth is each upstream issue — this is a tracker.

- **Collect run performance metrics**
- [ ] PennLINC/babs#356 — via con/duct resource monitoring (also tracked by #16)
- [ ] https://github.com/PennLINC/babs/issues/410 -- from SLURM
- [ ] ensure running `babs status` after array of jobs completes (e.g., within `mechababs iterate` before `merge` or upon `status`)
- file an issue to babs to allow for the "trailer" job with house-keeping to fully automate final collection
- [ ] follow up upon that `merge` -- pick up from not-committed `code/job_status.csv` within derivatives and commit as entries within `participants+jobs.tsv` ? check for congruency with con/duct collection
- **Scalability: Remote pull/push to avoid need for large store on compute**
- [ ] https://github.com/PennLINC/babs/issues/401 -- make RIAs remote
- potential issue: `babs merge` needs to run where we have babs status file and all the branches (not necessarily content)
- workout flow of "remotes" so we could push from ephemeral compute to "output_remote" which is on typhon (`origin` of the `origin` for the ephemeral clone; will be known):
- smaug/OpenNeuroStudies --(worktree or clone)--> working-copy-on-typhon (e.g. on typhon e.g by asmacdo) --> unity hpc for compute --> "ephemeral compute clone" (if it was a 'worktree' , remotes are *the same*! https://github.com/PennLINC/babs/issues/338; TODO: move into an issue)
- **Ensure re-rerunnability & portability**
- To be able to rerun at least a subject-specific run elsewhere, e.g. on typhon. Current showstoppers (TODO: move into issue(s)):
- absolute paths in the scripts
- leakage of SLURM env vars inside the scripts or datalad run records
- **BIDS-study organization**
- [x] PennLINC/babs#369 — fit BIDS-study layout (**merged**, `2cc536a`)
- [x] PennLINC/babs#392 — output folder `outputs/` → `derivatives/`; closed, superseded by #369's layout work
- [x] PennLINC/babs#372 — make `logs/` `.gitignore` optional — **delivered** by merged PR PennLINC/babs#386 (`babs init --no-ignore`)
- [x] PennLINC/babs#394 — allow full BIDS-acceptable chars in a derivative name (open PR)
- [ ] PennLINC/babs#329 — containers subdataset configurable path
- [ ] PennLINC/babs#378 — default `.gitattributes` BIDS-friendly (PR PennLINC/babs#393 open)
- **Avoid lengthy and potentially unnecessary zips**
- [ ] PennLINC/babs#327 — make zipping optional. Replaces the old PennLINC/babs#381 line: that PR closed without merging, and a minimal retry (skip the zip step only) is planned.
- **Use ///repronim/containers with `datalad containers-run` (or at least with our `scripts/singularity_cmd`)**
- [ ] TODO formalize TODO
- [ ] PennLINC/babs#383 — resolve container image path via datalad-containers; cf PennLINC/babs#329
- **Misc *babs* operational issues**
- [ ] PennLINC/babs#379 — add dataset license option
- [x] PennLINC/babs#385 — stage BIDS-inherited metadata for every job (open PR; carries @djarecka's PennLINC/babs#376 commits, which closed unmerged into this; closes PennLINC/babs#374; tracked by #41)
- [X] ~PennLINC/babs#384 — faithful BIDS-inheritance staging test. Follows PennLINC/babs#385, the inheritance-staging PR, which hasn't merged yet.~ cut, no hooks, but I have tested bids-inheritace
- [ ] PennLINC/babs#388 — modernize the slurm-docker-ci CI image off EOL CentOS 7 (tracked by #65)
- [ ] PennLINC/babs#395 — session-level `*prep` runs die at workflow construction; regression from PennLINC/babs#337 (open draft PR)