Protein-specific coordinates based on index groups

[jhenin]

The goal is to make the alpha_angles CVC (and dihedralPC) available on all back-ends.

TODO:

• [x] allow for series of groups to describe non-contiguous alpha segments
• [x] extend to dihedralPC
• [ ] update documentation
• [ ] regression & unit tests

[giacomofiorin]

A general comment: besides making those functions more portable there would be major advantages in making all CVCs rely on atom groups.

I think it would be great to just use the three groups groupCA, groupN and groupO consistently in all backends, and just have VMD and NAMD auto-fill them to preserve backward compatibility.

While working on #644 I realized that when that is done, CVCs that are based on centers of mass could easily be used any positions, including individual atoms. This would much simplify the data structures of the alpha and dihedPC.

[jhenin]

About making the groups groupCA, groupN and groupO generic: I realized that for the alpha CVC, we need to let the user specify several individual ranges of residues to track different alpha-helical segments. This is not compatible with having a single generic name for the index groups.

I don't get the comment about CVCs based on COMs being able to use individual atoms - isn't that already the case?