ENH: reference workflow 2.0

[eilidhmacnicol]

Changes proposed in this pull request

• remove multiple iterations of bold reference generations
• run once at start and pass outputs to rest of the functional processing workflow
• supercedes Reference workflow integration #35 and addresses Do not run artsBrainExtraction twice - fast BOLD reference masking #33

Unlike in #35, this reference workflow creates one reference image per BOLD run. In future, it would be good to implement something as in nipreps/nibabies#15 where reference images are calculated for similar BOLD runs (same PE, readout time) but this can be added at a later stage.

Documentation that should be reviewed

[eilidhmacnicol]

The checks that are failing are only doing so on the install in confined environment - [ setup.py - install ] workflow. This seems to be because of a dependency clash between the updated nipreps/niworkflows#656. I think the order of installation proceeds with tedana dependencies being installed first, and the newest version of matplotlib is being installed as a consequence of their non-specific dependency, see here, which later conflicts with niworkflows pinned dependency.

@oesteban - I don't know how to proceed with this, but the circleCI checks are looking good with the workflow completing without errors. WDYT?

[eilidhmacnicol]

@oesteban - I think these changes are good to go; can we address the pip build before merging?

[oesteban]

Okay, you should be able to rebase or merge upstream/master (don't forget to run git fetch upstream first)

[eilidhmacnicol]

The removal of nitime from dependencies (after rebasing to master) caused the last crash.
nitime is used in the confounds workflow by nipype, and since this is common to fmriprep but nitime is not included in its dependencies, I suspect that nitime is a dependency for a package required for fmriprep but not fmriprep-rodents. I tried digging around but it wasn't immediately obvious so adding it back in for the time being.

Fingers crossed that this time we get a clean sheet

[eilidhmacnicol]

Ding Ding Ding! We have a winner.

I've noticed that our ME pipeline might need some work before release (e.g. the masking is still using niworkflows rather than nirodents) but merging this and will test on SE and ME data next week. 🎉