Improve run-to-run reliability of confounds

[oesteban]

This is a meta-request for inclusion as part of the 20.1.0 LTS release, to reference issues that will help with the run-to-run reliability of confounds on surrogate projects (sMRIPrep, SDCFlows, NiWorkflows):

• BOLD reference/brainmask (niworkflows#485), which will require some additional testing (facilitated with niworkflows#488)
• Replace FSL FAST with FS' aseg whenever possible (smriprep#170)

[oesteban]

Both ideas are now implemented. Closing.