Some suggestions to make nuth_kaab and deramping more consistent with ICP results

[adehecq]

• made changes to deramping so it works with masked Raster arrays
• added a function to calculate a rotation, translation between two PCs. This was initially thought as a way to convert a "Ramp" into a transformation matrix (which seems to work!), but I'm curious to see how it performs on real point-clouds.
  @erikmannerfelt if you have a chance to take a look, I'm curious to hear what you think?

[rhugonnet]

@erikmannerfelt @adehecq where are we on this, and in relation to #87? I'll dive into the details only if it is still needed

[erikmannerfelt]

@rhugonnet the small corrections have been implemented in different PRs and I am working on the superinpose implementation with the Coreg subclass (see #94). Therefore, I will close this PR.