relax SVN archive

Fix for the convergence tests in full_analysis.py.

The spin name can now be absent.

Modified the convergence tests in the full_analysis.py script to print the IEEE-754 byte array.

This representation is platform and float type (float vs. numpy.float64 vs. Numeric.Float64, etc.)
independent. Even if the values evaluate to being equal, the python print out can be different.
The IEEE-754 float byte array representation is independent of this insanity.

The IEEE-754 byte arrays are now restored from the results files, giving full precision.

Modified the XML results files to store higher precision versions of the floats.

The float is stored as an IEEE-754 byte array in the XML attribute ieee_754_byte_array.

Added the reset_min_stats() method to reset all minimisation stats prior to optimisation.

Fixes for the model_selection() method of the full_analysis.py script.

The model selection pipe must be deleted each time for the automatic CONV_LOOP looping to be
possible.

Fix for the full_analysis.py script. The references to the diffusion tensor object were incorrect.

Another fix for the duplicate_data() method.

The diffusion tensor object was being copied from the source pipe back to the source pipe!

Fixes for the duplicate_data() method.

The diffusion tensor object now has its sub-objects compared.

Fix for the OMP model-free optimisation system test.

The 'aic' data pipe already exists, so now it is deleted prior to model selection.