xfm2tag unsafely assumes image and mask coordinates are the same

[bcdarwin]

See here and here.

Arguably the real culprit here is C: get_volume_real_value exhibits undefined behaviour when fed garbage voxel indices. I found this bug in a pipeline where xfm2tag was segfaulting, but the same user has used the same combination of initial model (which determines the resolution of the output images) and different-resolution mask (which is used in the xfm2tag call) before and didn't get any errors, although it's possible tag points were not generated as a result ...

[gdevenyi]

I think I got bitten before by different resolutions of masks and atlases.