Revise SBRef handling

[oesteban]

• Correctly handle multiple 3D and 4D SBRefs (see N4BiasFieldCorrection error in sbref image #1463). That implies agreeing on a policy for SBRefs (relates to Use subject specific BOLD template for T1w coregistration #450).
• Address differences in resolution (initiated in [WIP] resample sbref niworkflows#253).
• Better boilerplate (Update boilerplate to correctly handle single-band references niworkflows#413)

[oesteban]

I'm pushing the priority of this up. In working through #1803, I've realized the workflow for SBRefs needs an urgent revision.