relax SVN archive

Fix for bug #18030 (https://gna.org/bugs/?18030).

The problem was that the grid search limits were being scaled iteratively for each successive spin
system. This only affects certain type of custom grid searches.

The grace file created by pcs.corr_plot() now separates each element into its own graph.

Bug fix for the N-state model for when only the paramagnetic centre position is optimised.

If the alignment tensor is fixed, as well as the probabilities of the states, then the optimisation
of the paramagnetic centre position was failing as the wrong target function was being used and the
correct one was not set up correctly for this.

Expanded the N-state model test_align_fit() system test to check the pcs.back_calc() user function.

The RDC and PCS Q factor user functions now do nothing when data is missing rather than failing.

A warning is given and the function now simply returns rather than crashing.

Fix for the new create_frq_label() function.

Added a new relaxation data function create_frq_label().

This is for generating a frequency label in MHz rounded to the nearest factor of 10.

Fix for the conversion of te to ts - the model is not always given.

The te parameter is now properly converted to ts is the model is one of m5, m6, m7, or m8.

Removed 'relax_disp' from the VALID_TYPES data pipe type array.

This is supported in the 'relax_disp' branch and not in the main line.