relax SVN archive

Converted the frame order test data distribution generation script to the interatomic data design.

Converted a number of objects in the frame order target functions to numpy.float16 types.

These were originally float64. These objects relate to the 3D structure which is not very accurate
anyway. This should help with the movement of data for the multi-processor optimisations.

The rigid frame order model no longer uses the numerical quasi-random integration path.

The free-rotor pseudo-ellipse frame order target function is now using the multi-processor framework.

This is for the func_pseudo_ellipse_free_rotor_qrint() target function.

Converted all of the CaM frame order test data scripts to the interatomic data design.

Fixes for all of the RDC and PCS flags in the frame order target functions.

Converted the frame order target function code to be more like the N-state model.

This matches a number of changes which occurred in the trunk (via branches). The alignment index
has been changed from 'i' to 'align_index' for clarity. The 'rdc_const' and 'pcs_atoms' keyword
arguments have been renamed to 'dip_const' and 'atomic_pos' respectively to match the N-state model.
And proper support for the interatomic data design for the RDCs has been added.

The frame order system tests are now showing the number of integration points as a debug message.

Fix for the frame order _minimise_setup_rdcs() method for handling multiple unit vectors.

Fix for the Frame order _minimise_setup_pcs() method for how atomic positions are now stored.

This will always be a rank-2 array, even with a single position, and the method now handles this.