FIX: Coerce MGHImages to Nifti1Images when plotting

[mgxd]

Supercedes nipreps/smriprep#317

[mgxd]

This fixed the errors previously encountered in the nibabies runs
https://app.circleci.com/pipelines/github/nipreps/nibabies/383/workflows/d14784ad-3d72-4ed4-a860-a5b0a357bac9/jobs/1388

[effigies]

We should report this upstream to nilearn, as well. Workaround is fine because all we care about is data/affine, but they at least used to be able to handle SpatialImages in general, so this seems like a regression.

[mgxd]

@effigies - Looks like it needs to be a NiftiImage, so might as well coerce.

https://nilearn.github.io/stable/modules/generated/nilearn.plotting.plot_anat.html#nilearn.plotting.plot_anat
https://nilearn.github.io/stable/manipulating_images/input_output.html#inputing-data-file-names-or-image-objects

[effigies]

https://nilearn.github.io/stable/manipulating_images/input_output.html#niimg-like-objects

A string or pathlib.Path object with a file path to a Nifti or Analyse image
An SpatialImage from nibabel, ie an object exposing get_fdata() method and affine attribute, typically a Nifti1Image from nibabel.

[effigies]

I'm +1 for merging rather than relying on people not to have this specific version of nilearn.