Difference in INU behavior when providing precomputed mask

[psadil]

Describe the bug
In a typical run without a precomputed mask, INU is done on the whole-head image, with weights that can emphasize WM and the brainstem. In a run with a precomputed mask, INU appears to be done on the masked image, and the weights differ.

I think this happens in this branch:

smriprep/src/smriprep/workflows/anatomical.py

Lines 920 to 926 in 3c76b5b

	# Run N4 if it hasn't been pre-run
	if not have_t1w:
	LOGGER.info('ANAT Skipping skull-strip, INU-correction only')
	n4_only_wf = init_n4_only_wf(
	omp_nthreads=omp_nthreads,
	atropos_use_random_seed=not skull_strip_fixed_seed,
	)

init_n4_only_wf doesn't have a way to pass wm_prior to init_atropos_wf https://github.com/nipreps/niworkflows/blob/d991644f3a769d7c84e7c8bb897ea30fb96885df/niworkflows/anat/ants.py#L961-L966

so wm_prior always takes the default value of False https://github.com/nipreps/niworkflows/blob/d991644f3a769d7c84e7c8bb897ea30fb96885df/niworkflows/anat/ants.py#L474

meaning that the wm_prior is not applied https://github.com/nipreps/niworkflows/blob/d991644f3a769d7c84e7c8bb897ea30fb96885df/niworkflows/anat/ants.py#L793-L822

I think this difference in use of weights is a bug, but you may disagree. At minimum, I think this is behavior that should be documented (here and in fmriprep).

Exact command line executed

❯ docker run --rm -it -v /Users/psadil/Desktop/mask-test:/Users/psadil/Desktop/mask-test nipreps/smriprep --fs-license-file=/Users/psadil/Desktop/mask-test/code/license.txt --derivatives=/Users/psadil/Desktop/mask-test/sourcedata/synthstrip $PWD/sourcedata/raw $PWD participant

# or
❯ docker run --rm -it -v /Users/psadil/Desktop/mask-test:/Users/psadil/Desktop/mask-test nipreps/smriprep --fs-license-file=/Users/psadil/Desktop/mask-test/code/license.txt $PWD/sourcedata/raw $PWD participant

Are you positive that the input dataset is BIDS-compliant?

• I have used the online BIDS-Validator
• I have run a local installation of the BIDS-Validator (please, note the version of the validator here).
• I let sMRIPrep check it for me (in other words, I didn't set the --skip-bids-validation argument).
• No, I haven't checked myself AND used the --skip-bids-validation argument.

sMRIPrep feedback information

Visual reports of sMRIPrep
This is unrelated to the main concern (differences in how INU is applied), but the resulting reports end up looking different; the outputs with a precomputed mask have the mask applied.

Without precomputed mask

With precomputed mask

Installation type (please complete the following information):

• "Bare-metal"
• Singularity
• Docker

Additional context
Possibly related: nipreps/nibabies#397