relax SVN archive

The relaxation dispersion calculate user function now stores the back calculated R2eff values.

A number of changes were required for this. The code from the end of the Disp_result_command.run()
method was converted to the function specific_analyses.relax_disp.disp_data.pack_back_calc_r2eff().
This allows the back calculation R2eff unpacking code to be shared. The new has_proton_mmq_cpmg()
function has also been created to simplify the code.

Bug fix for the multiple quantum relaxation dispersion models.

These require both the heteronuclear and proton chemical shift differences. But the proton
difference was being scaled by the heteronuclear Larmor frequency and not the proton frequency.

Created a Grace plot of the 15N MQ CPMG data fitting from cpmg_fit.

Created the Relax_disp.test_korzhnev_2005_all_data system test for checking the 'MMQ 2-site' model.

This checks against all six data types, 1H SQ, 15N SQ, DQ, ZQ, 1H MQ, and 15N MQ. This is currently
set to the values found by cpmg_fit. As this is the true solution, relax should find similar
parameter values.

Bug fix for the specific_analyses.relax_disp.disp_data.find_intensity_keys() function.

This function was not handling multiple experiment types correctly.

Improvement for the file names in the relax_disp.plot_disp_curves user function.

The '_' character is now used between the experiment name and the rest of the file name.

Updated the Relax_disp.test_korzhnev_2005_15n_dq_data system test so it passes.

The optimised values are very similar to that from cpmg_fit, so the code must be functioning
correctly.

Better support for the MMQ-type data dispersion models for the end of the optimisation.

The back calculated R2eff values are now handled correctly for the attached proton in the spin
system.

Updated the relax script for optimising Flemming Hansen's CPMG data.

Fix for the previous commit (r21562).