ENH: Improve support for infant MRI templates

As part of https://github.com/mne-tools/mne-bids-pipeline/issues/574, I've created a single subject with a mne coreg warp of the ANTS6-0Months3T infant template provided by fetch_infant_template. There is currently no "nice" way for me to include this AFAIK in the dataset. I see two options (maybe we want to support both?):

1. Include as freesurfer derivative

I could add the warped MRI as derivatives/freesurfer/<subject>, but mne-bids-pipeline doesn't seem to have the logic to look for already-existing BEM solutions. I could add this support and upload to openneuro with the warped MRI as a FreeSurfer derivative, then I think everything could "just work".

2. Add support for non-`fsaverage` templates

This would be a potentially cleaner solution since it would only require me shipping one small text file rather than a whole FreeSurfer subject directory. It would require something like

[ ] Add support for non-fsaverage templates in MNE-BIDS-Pipeline (at least the ANTS* templates).
[ ] Include in our BIDS derivatives (somewhere?) the -trans.fif plus the MRI scaling parameters.cfg file that tells mne.coreg.scale_mri how to scale ANTS6-0Months3T to make the template subject. This could be done in a backward compatible way for fsaverage where, if not found, uses the current trans (and uniform scaling?) method.

From this the necessary FreeSurfer subjects could be created.

Notably, in either of these cases, the BEM code would need to be updated a bit -- the data are MEG-only, which triggers MNE-BIDS-Pipeline to want to use 1-layer BEMs, but the ones for ANTS* templates are all 3-layer (and I think it should be okay/acceptable to use 3-layer BEMs with MEG data if you want and they are already available).

We could have a quick video chat about this at some point to hash out the ideas @hoechenberger @agramfort

mne-tools / mne-bids-pipeline