relax SVN archive

The Ulysses butterfly is now the symbol of relax.

The manual cover has been touched up an the ulysses image added.

Started to insert the Ulysses picture at the cover of the manual and removal of the non-eps
dependencies images as they are stored in the relax paper.

Small change to the full_analysis.py sample script.

Bug fix:

Another large bug fix! This time the linear constraints were failing due to the diagonal scaling.
The tm, te, ts, and tf parameters must all be scaled the same if a constraint with more than one
of these parameters is used.

Massive bug fix!!!

Caught the bug causing problems with Newton minimisation of the diffusion parameters together with
the model-free parameters. The scaling matrix was not being constructed properly! Only the first
residue's parameters were being scaled 'index = 0', all the rest were unscaled. The bug was because
the model-free specific minimise function was passing in the index of zero rather than None!!!

Changed the global default line search algorithm from More and Thuente to Backtracking.

Monte Carlo simulations in the full analysis sample script have been fixed up.

Previously each simulation was a global optimisation, now only the model-free parameters are being
optimised. Otherwise the error analysis on a full protein would take way too long (feedback from
Chris MacRaild). The errors are now slightly underestimated as the diffusion tensor has no errors,
but the speed is much more important.

Added a hint to the full analysis sample script.

Bug fix:

Fixed a bug which manifested itself in the full analysis sample script when optimising the diffusion
tensor after reading in the local tm results. The bug was in the add_residue function of the
relaxation data specific functions class and was because the 'data.noe_r1_table' data structure was
not being properly created.

Really fixed a bug in the full analysis sample script this time.