ENH: create subworkflow for defacing of PET data without an MRI

[mnoergaard]

In cases where no anatomical MRI as acquired in combination with the PET data (happens quite often for clinical PET data), it would be beneficial to have a workflow that can still do adequate defacing. This can potentially be solved by 1) doing defacing of a template anatomical brain, 2) co-registration (12 parameters) between PET and MRI, and 3) defacing of the PET data. Ideally this workflow should automatically run if no MRI data is detected for a given subject.

[bendhouseart]

As a brief update, I did successfully run petdeface with a an averaged PET image cast as a BIDS T1w. I know that's a little bit different than initially discussed, but if we can make it work that way it solves registration issues quite handily, otherwise happy to try out a template.