ENH: Refactor workflow to adopt *SDCFlows*' new correction workflow

[oesteban]

This PR addresses the problem reported in #2210.

Along with the specific changes in SDCFlows, this PR:

• Resamples BOLD data in its native space using new SDCFlows' utilities.
• Introduces a new IdentityInterface node called bold_final that acts as buffer to reduce LoC and branches when dealing with data with/without fieldmaps (and helpful too for ME-EPI, although this is not fully implemented).
• ROIs plot now uses the output of unwarping through a RobustAverage node, therefore fixing one issue reported in Susceptibility distortion correction (SDC) using Spin Echo Fieldmaps  #2210.
• preproc datasink of BOLD in native space now connected to the bold_final node, which contains HMC'ed + SDC'ed (if fieldmaps found) BOLD in native space. Now, it is matched with the ROIs plot, reducing a little the opportunities for divergence.

This PR DOES NOT solve (or even leaves without implementation):

• Resampling into T1w and standard spaces when fieldmap is present - the same error for the bold data reported in Susceptibility distortion correction (SDC) using Spin Echo Fieldmaps  #2210 is expected to happen, as the displacements field generated by SDCFlows does not look right all times (B0FieldTransform botches the displacements field with flipped axes sdcflows#236)
• Mapping of surfaces into EPI
• ME-EPI + fieldmaps (still trying to resolve problems with single-echo, will look into this at a later time).

[oesteban]

@mgxd I've rebased on master after running black on workflows/bold/base.py, so the diff should be much cleaner now. Some little stylistic changes remained, but now they are just residual changes I don't think will disturb the review process.

[oesteban]

This is holding me up a little, so I'll merge.