Add neuroimaging packages

[joschaschmiedt]

Follow-up of #3

• [x] nipype
• [x] dipy
• [ ] mindboogle
• [x] nibabel
• [x] napari-nibabel
• [ ] scitran SDM
• [ ] nipy
• [x] nibabel
• [ ] Scitran SDM
• [ ] nipy
• [x] nitime
• [x] popeye
• [x] nilearn
• [ ] pymvpa
• [x] mne
• [x] mnelab
• [x] mne-realtime
• [x] mne-rsa
• [x] mne-nirs
• [x] mne-icalabel
• [x] mne-connectivity
• [x] mne-features
• [x] mne-bids
• [x] mne-bids-pipeline
• [x] mne-ari
• [ ] niwidgets
• [x] pybids
• [ ] nistats
• [x] imageio
• [x] pydicom
• [x] deid
• [x] intensity-normalization

[pantaray]

I've included a first batch of neuroimaging packages in the environment. Some hiccups:

• mindboogle is not packaged per se but only provides a Docker image
• scitran does not provide a package, only a REST API
• nipy: problematic on macOS and Windows. The conda package only supports Python 3.8, pip does not provide a wheel and needs to build nipy against BLAS/LAPACK which currently crashes on both macOS and Windows.
• pymvpa: from the docs "...Python 3.X should work too, but none of the core developers are using it in production (yet), hence it should be considered as less tested...."
• niwidgets: depends on scikit-learn < 0.21 (we're using 1.2.1)
• nistats: dependency conflict with nibabel/intensity-normalization

I think not providing nipy on all platforms might be the biggest omission here. However, that still strongly depends the actual application. @joschaschmiedt Do we want to proceed with nipy only on Linux and see if and how many macOS/Windows users actually need it? Tests with the current env file are passing: https://github.com/neuro-conda/neuro-conda/actions/runs/4344547143

[joschaschmiedt]

Thanks! Supporting Linux only is better than not at all, so I think that's fine. About the other issues, there's not much we can do about that, right?

See also my comments over at #29

[pantaray]

Yes, aside from actively patching/rewriting things I'm afraid we can't do much here.