ENH: Large overhaul of the functional workflow w/focus on ME-EPI

[oesteban]

This PR refactors MRIQC to handle ME-EPI better. Incidentally, it also includes brain extraction (on the first echo, or single echo) with SynthStrip.

Specific changes:

• Head motion is estimated on the first echo
• Non-steady states are estimated on the second echo (the first has too much T1 contribution).
• Synthstrip is used in brain extraction (first echo or single echo).
• HMC is done with nitransforms, applying the estimation on first echo to all of them.
• Single report for all echos.

Resolves: #773.
Resolves: #1152.
Resolves: #1028.

[oesteban]

Seems to be performing as expected, both on circle (all of them se-epi) and locally (me-epi).

[effigies]

LGTM. Mainly checked the new SelectEcho interface and the synthstrip workflow. Most of the rest looks like necessary plumbing, but it would take me more time than it's worth to acquaint myself with the workflow and verify each one. If the workflow is running, that seems like a good sign that they worked.

[oesteban]

Mainly checked the new SelectEcho interface and the synthstrip workflow.

Yes, those are the only things with reuse potential from fMRIPrep, and candidates for upsourcing into niworkflows.