make discretization self-contained

[franknoe]

Currently, the Ramachandran basis set files contain eigenvectors values over phi/psi. The knowledge how the discretization has been made (10 degree bins from -180 to 180) is in the code reading them. This is not good for extensibility because a person that wants to add new parameters now has to back-engineer the code and format their files accordingly.

It would be better to save the discretization / binning information (xbins, ybins) together with the data. In this way, the code is also compatible with data files that use different discretizations.

In order to sort the data into bins, consider using np.digitize.

This issue is not urgent.