SDC - Horrible co-registration of *b=0* reference and EPI reference

[oesteban]

The inputs to antsRegistration when finding the misalignment between the fieldmap reference and the target DWI (via its b=0 reference) is currently unacceptable for our dataset ds001771.

This problem is very likely to be present in any dataset, as the b=0 reference we are using right now (work/dmriprep_wf/single_subject_36_wf/dwi_preproc_acq_AP_wf/dwi_reference_wf/enhance_and_skullstrip_dwi_wf/normalize/normalized.nii.gz) is definitely suboptimal.

Perhaps, we should move towards the new reference workflow in niworkflows, if we haven't done so yet.

[oesteban]

Seems to be fixed in #169 with production settings, but still problematic for ds001771 in debug settings.