relax SVN archive

Ported r2646 to r2647 from the 1.2 line.

The command used was:
$ svn merge -r2645:2647 svn+ssh://bugman@.../svn/relax/1.2

This adds a number of model-free optimisation tests to the test suite.

Ported r2643 from the 1.2 line.

The command used was:
$ svn merge -r2642:2643 svn+ssh://bugman@.../svn/relax/1.2

This includes a few small formatting changes to the model-free file.

Ported r2642 and r2644 of the 1.2 line (r2643 is not included).

The commands used were:
$ svn merge -r2641:2642 svn+ssh://bugman@.../svn/relax/1.2
$ svn merge -r2643:2644 svn+ssh://bugman@.../svn/relax/1.2

These changes include improvements to the infrastructure chapter of the relax user manual, and a
large expansion of the relax development chapter.

Ported r2641 from the 1.2 line.

The command used was:
$ svn merge -r2640:2641 svn+ssh://bugman@.../svn/relax/1.2

This fixes documentation bug #7402 (https://gna.org/bugs/?7402).

Ported r2637 to r2640 from the 1.2 line.

The command used was:
$ svn merge -r2636:2640 svn+ssh://bugman@.../svn/relax/1.2

This includes a number of changes to the Scons system including fixes for the distribution target
functions, the creation of the new module 'scons.install', and a restructure of the base
'sconstruct' file.

Ported r2635 from the 1.2 line.

The command used was:
$ svn merge -r2634:2635 svn+ssh://bugman@.../svn/relax/1.2

This involves a few functions moving to the new Scons module file 'scons/distrib.py'.

A fix for the function which loads the sequence from the PDB file.

This fixes two tests of the test suite. The changes already exist within the 1.3 line.

Speed ups for the optimisation tests in the test suite.

The grid search of each optimisation algorithm test has been replaced by the value setting function
as the results of the grid search are the same for each of the tests. As the grid search is
computationally expensive, this significantly speeds up the tests.

Added a number of model-free optimisation tests to the test-suite.

These tests use the synthetic data back calculated for the model-free parameter values of {S2=0.970,
te=2048, Rex=0.149}.

The following optimisation techniques are currently tested:
Grid search.
Steepest descent, backtracking line search.
Steepest descent, More and Thuente line search.
Coordinate descent, backtracking line search.
Coordinate descent, More and Thuente line search.
BFGS, backtracking line search.
BFGS, More and Thuente line search.
Newton, GMW Hessian modification, backtracking line search.
Newton, GMW Hessian modification, More and Thuente line search.

Fixes for a number of bugs caused by the changes for task #3838.

Task #3838 is located at https://gna.org/task/?3838 and more information is given at
https://mail.gna.org/public/relax-users/2006-10/msg00002.html (Message-id:
<7f080ed10610032153o4571935bq138d2739f55ca3bd@...>).

Part of the problem, the failure of two of the tests of the test suite was reported by Alex Hansen
(viochemist _at_ gmail _dot_ com) in the post located at
https://mail.gna.org/public/relax-devel/2006-10/msg00095.html (Message-id:
<481156b20610130720u39a93e30k6d4603efd8730c96@...>). Another failure was reported by
Alex at https://mail.gna.org/public/relax-devel/2006-10/msg00096.html (Message-id:
<481156b20610130728jb6aa76ag3d9487a20397391b@...).

The first fix, which was a missing run argument, is described at
https://mail.gna.org/public/relax-devel/2006-10/msg00099.html (Message-id:
<7f080ed10610130741j146721c8w512ba61bac3644c7@...>).

The second problem was that the residue number and residue name was being referred to as 'num' and
'name' when they should have been 'res_num' and 'res_name'.

The third problem was an index for selecting the residue. Rather than using the real index 'i', the
index 'j' was being used.