ENH: Add label iamge registration and tests.

[ntustison]

@stnava , @cookpa --- I was going to send an email regarding this work but figured it would be easiest to put all the information in a possible pull request.

With the segmentation networks developed over at ANTsXNet, I've been leveraging that information for registration and I figured the functionality might be useful for the community. Basic usage takes respective sets of fixed and label images, converts them to sets of single label images and calculates the centers-of-mass of all the labels. Those centers-of-mass are used to quickly compute a linear transform using ants.fit_transform_to_paired_points. The single label images (and optional intensity images) are used as input to one of the antsRegistrationSyN* deformable-only calls. I've also been playing with this for masked lesion registration.

Also, add test.

Let me know if you have a problem with this pull request. Alternatively, I can put it over in ANTsPyNet.

[coveralls]

coverage: 85.003% (-0.06%) from 85.061%
when pulling 76857fc on LabelReg2
into 96ba4f7 on master.

[stnava]

this is great! been on the to do list for a long time.

[ntustison]

Thanks @stnava . I'll merge this in just a sec and then I'll work on the tutorial example and implementing the ANTsR analog.

[ntustison]

Self-contained example