ENH: Reorient interface

[effigies]

This is a pure nibabel/numpy interface to reorient an image to any RAS ... IPL. It also produces an affine transform that describes the reorientation.

Based off of fmriprep.interfaces.images.Reorient, but does not assume RAS target orientation or nibabel >= 2.2.

I created a interfaces.image file, since this didn't seem to fall into existing interface categories. Is there a better place to put this?

[codecov-io]

Codecov Report

Merging #2572 into master will increase coverage by 0.02%.
The diff coverage is 90.19%.

@@            Coverage Diff             @@
##           master    #2572      +/-   ##
==========================================
+ Coverage   67.57%   67.59%   +0.02%     
==========================================
  Files         333      334       +1     
  Lines       42547    42598      +51     
  Branches     5266     5273       +7     
==========================================
+ Hits        28750    28796      +46     
- Misses      13124    13126       +2     
- Partials      673      676       +3

Flag	Coverage Δ
#smoketests	50.7% <ø> (ø)	⬆️
#unittests	65.06% <90.19%> (+0.05%)	⬆️

Impacted Files	Coverage Δ
nipype/interfaces/image.py	90.19% <90.19%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5a96ea5...ea5a5e2. Read the comment docs.

[oesteban]

+1000, and I think the new module is fine.

[effigies]

Well, hoped to get #2576 to test the nibabel fallback code, but that's not likely to happen.

This should be good to go.

[oesteban]

Seems like some tests are failing in Travis. Particularly, when checking that the affine was actually changed. Seems like (0, 1, 2) affines are being set?

[effigies]

Ah, no. This is the case where there's a flip but not a reordering. The dimension info doesn't include direction, just axis.

[effigies]

@oesteban Good to merge?