Potential computation savings for Build-Model/MAGeT

[cfhammill]

Just discussed briefly with Ben - it may be possible to use the transforms from the build model to the perform the second step of MAGeT very cheaply. The atlas->template step would be needed but
instead of template->subject we could use concatenation of template->average, average->subject from the build model. Would need to be validated of course.

[bcdarwin]

Pydpiper contains some support for this but the results weren't as good so I didn't switch over to using it, but this may be a minctracc vs ANTS issue; see https://www.biorxiv.org/content/early/2018/03/06/277087 ...