relax SVN archive

Renamed the Numeric matrix 'select_sim' to 'all_select_sim'.

This matrix contains the Monte Carlo simulation selection status of all spin systems of all
simulations. Hence the name 'all_select_sim' has been chosen to differentiate it from the array
'select_sim' stored at 'self.relax.data.res[run]'.

Fix for bug #7755 (https://gna.org/bugs/?7755).

This bug was reported by Stephen Headey (https://gna.org/users/sjheadey).

The problem was that the select_sim array created by the columnar results reading function
'self.read_columnar_results()' in 'specific_fns/model_free.py' was one massive single array
containing all the selected simulation data of all spin systems of all simulations. Subsequently
this was then not being split up for each spin system. Hence the total number of simulations was
being set to the number of spin systems times the number of simulations. The select_sim array for
each spin system was also of the same length and contained the data for all spins.

The fix was to convert the 'select_sim' array being created by 'self.read_columnar_results()' into
a Numeric matrix. The first dimension corresponds to the simulation and the second corresponds to
the instance.

Added print out messages for when spin systems are being eliminated.

Ported r2891 from the 1.3 line.

The command used was:
svn merge -r2890:2891 svn+ssh://bugman@.../svn/relax/1.3

This fixes bug #7875 (https://gna.org/bugs/index.php?7875).

Fix for bug #7875 (https://gna.org/bugs/index.php?7875).

This bug was reported by Sebastien Morin (https://gna.org/users/semor).

The problem was that the 'model_selection()' user function was incorrectly copying the data into the
model selection run when the local tm global diffusion model was selected over the sphere, spheroid,
ellipsoid, etc. The local tm global diffusion model was being treated as a single spin system model
by the 'self.duplicate_data()' function of 'specific_fns/model_free.py' and hence only the data from
the first spin was being duplicated. By passing the 'global_data' flag in from the model selection
code, the data duplication function is now acting correctly.

initial commit of maths fns unit tests and test for the unit test framework

note: all tests are just dummy null tests (including failures)

Ported r2888 from the 1.3 line.

The command used was:
svn merge -r2887:2888 svn+ssh://bugman@.../svn/relax/1.3

This fixes bug #7863 (https://gna.org/bugs/index.php?7863).

Fix for bug #7863 (https://gna.org/bugs/index.php?7863).

This bug was reported by Sebastien Morin (https://gna.org/users/semor).

The fix was simply to set 'self.runs' to all the global diffusion models to be used by the AIC model
selection step.

Ported r2886 from the 1.3 line.

The command used was:
svn merge -r2885:2886 svn+ssh://bugman@.../svn/relax/1.3

This removes the PDF manual from the repository.

Removed the PDF user manual from the repository.

As the file is a large (2.6 Mb) binary file and changes often, it has been removed from the
repository. This should stop the relax Subversion repository getting too bign. The manual can
easily be compiled from the sources using Scons.