relax SVN archive

Fix for many of the frame order target functions.

The Rx2_eigen matrix is a local, not class variable.

Created the pseudo-ellipse frame order test model data.

The Monte Carlo integration flag is now stored at cdp.mcint and is not set for the rotor model.

This allows the method to be programmatically changed.

A few small bug fixes for the old quadratic integration target functions.

Target and auxiliary functions are now available for both quadratic and Monte Carlo integration.

This can be set via the mcint flag to the maths_fns.frame_order initialisation code. In some cases
quadratic integration is faster or better, whereas in others the rough Monte Carlo technique is
better.

Redesign of the numerical integration method for an ~40 times speed up.

This is currently only for the isotropic cone model. A basic and rough Monte Carlo integration
method has been hand coded to replace the advanced scipy.integrate.tplquad() quadratic integration
function. Not only are the number of integration steps massively decreased, but all PCS values can
be integrated simultaneously. I.e. there is one iteration of the integration algorithm for all PCS
values of all atoms and alignments, rather than one iteration per PCS.

Added the cProfile dependency check.

Enabled the cProfile module when available to massively speed up the profiling.

Updates to the isotropic cone frame order test model data.

Fix for the calculation of the PCS value via numerical integration for cone frame order models.

The surface normalisation factor sin(theta) was missing!