relax SVN archive

Updated all of the relax vs. cpmg_fit comparison files in the shared data directory.

These now show the perfect match between the programs. The cpmg_fit source code was modified to
improve the accuracy of the gyromagnetic ratio values.

Big bug fixes for the 'MMQ 2-site' dispersion model.

The equations in Korzhnev et al., 2004 are incorrect! This is clear when comparing the equations in
the paper to those implemented in the cpmg_fit source code. The equations in relax have been
changed to match the correct one.

In addition, the matrix power factor must be found with the python math.floor() function and not
int() as the later will sometimes round up.

Created a relax script to compare the 'MQ CR72' dispersion model results to cpmg_fit.

The cpmg_fit solution is used as the input parameters for relax, and then a calc user function call
is used to back calculate the R2eff values. These values are then plotted to show the perfect
match.

Simplified the first 'MQ CR72' dispersion model formula in the manual.

Last fix for the 'MQ CR72' dispersion model.

The wrong value was being subtracted from the first eigenvalue - the value of log(Q)/relax_T should
not be divided by the number of CPMG blocks.

Changed the sign of the delta_omega frequency for the ZQ data in the 'CR72' and 'MQ CR72' models.

General improvement for the optimisation of many target functions.

For those models which use the tau_CPMG value, this is now recalculated. This means that if a user
inputs truncated nu_CPMG values, these are corrected when calculating tau_CPMG so that full precision
values will be used for the optimisation.

Improvements for the 'CR72' and 'MQ CR72' dispersion model R2eff calculating functions.

The numpy.arccosh() function can handle all input values when complex, therefore the checks for the
real part being above 1 are not necessary.

Fix for the 'MQ CR72' dispersion model target function.

The correct R20 values are now extracted from the parameter vector.

Removed a latent bug in the 'MMQ 2-site' dispersion model.

This was not being seen but might have caused problems in the future.