relax SVN archive

Fixed the loops in the two C functions to go to '(int)num_times' rather than the temp value of '6'.

The 'back_calc_I' C function now works by returning a Python list instead of a Numeric array!

An attempt at making the 'back_calc_I' C function return a Numeric array.

Completion of the C code for the fitting of relaxation curves and preparation for Monte Carlo sims.

The 'chi2()' C function is now declared in the 'func()' function as returning a double. Previously
it was interpreting it the result as an int!

The exponential back calculation function has been debugged. The function is I0 * exp(-t * Rx) and
not I0 * exp(-t / Rx)!

In 'maths_fns/c_chi2.c' the function 'square()' has been defined.

The makefile 'maths_fns/Makefile' has been modified so that changes to 'maths_fns/relax_fit.h' will
now cause recompilation of all C code.

The spectral density function 'num_instances()' in 'specific_fns/jw_mapping.py' has been debugged.

In 'specific_fns/relax_fit.py', the function 'ave_and_sd()' has been renamed to 'mean_and_error()'.
The user function 'mean_and_error()' has been added to 'prompt/relax_fit.py'. Previously this was
being called by the minimise function of 'specific_fns/relax_fit.py', ie both during the grid search
and the minimisation. This function has to be called prior to optimisation.

For Monte Carlo simulations, the function 'back_calc_I()' has been added to 'maths_fns/relax_fit.c'.
This simply converts the C external variable 'back_calc[]' to a Numeric array and returns it. The
code isn't fully functional yet. In addition, in the file 'specific_fns/relax_fit.py', the
functions 'num_instances()', 'return_data()', 'set_selected_sim()', and 'sim_pack_data()' have been
added for MC sims.

The error function 'RelaxNoMfModelError' has been changed to 'RelaxNoModelError' to generalise it
for both model-free analysis and relaxation curve fitting.

The function 'sim_set_selected()' has been renamed to 'set_selected_sim()' in
'specific_fns/model_free.py'.

Fix bug where Ri values of None were read from results files

Added a catch for zero Rx values.

Fixed the chi-squared function and 'func' now calls it.

Assigned the back calculated intensity to the external array 'back_calc'.

Added some sanity to the PyArrayObjects in the C code.

Pointers, eg 'relax_times' have been created to 'numpy_relax_times->data' and recast as double. Now
the elements of the array are refered to as 'relax_times[2]' rather than
'relax_times->data + 2*relax_times->strides[0]'. This should remove much pain.

Improved the makefile rules for compiling the C code.