relax SVN archive

Constraints are now allowed for the paramagnetic position optimisation in the N-state model.

This is because the gradients are implemented and functional.

Bug fix for the calculation of the PCS gradient with respect to the paramagnetic coordinates.

This is in the ave_pcs_tensor_ddeltaij_dc() function - the gradient needs to be converted back to
the Angstrom units of the paramagnetic position parameters.

Comment fixes in the ave_pcs_tensor_ddeltaij_dc() and pcs_constant_grad() functions.

Bug fix for the N-state model optimisation of populations and paramagnetic position.

The gradient from dfunc_standard() is now correctly calculated when both are optimised.

Bug fix for the N-state model target functions - now copies of the gradient and Hessian are returned.

Expanded the N-state model target function func_standard() docstring to include the xi derivative.

This is the partial derivative with respect to the paramagnetic centre position.

The probabilities and paramagnetic position can now be simultaneously in the N-state model.

Such a combination would have previously failed to unpack the parameters correctly.

Added a RelaxError to the N-state model Hessian for the optimisation of the paramagnetic position.

This is because these equations are not derived or coded yet.

Removed the unused parameter scaling in the N-state model gradient and Hessian target functions.

Major code simplification of the N-state model target functions.

The func_tensor_opt(), dfunc_tensor_opt(), and d2func_tensor_opt() methods have been merged with the
func_population(), dfunc_population(), and d2func_population() methods into the new func_standard(),
dfunc_standard(), and d2func_standard() methods. This halves the amount of code required to be
maintained and debugged. For the merger, the new probs_fixed class instance variable has been
created to determine when the probabilities need to be unpacked from the parameter vector.