relax SVN archive

Fix for the test_suite.clean_up.deletion() method for another MS Windows problem.

Sometimes the failed rmtree() call actually deletes the files and throws the WindowsError error.
Therefore the second rmtree() call will throw another WindowsError for the missing files. This is
now caught.

Fix for the relaxation dispersion system test tearDown() method.

The rmtree function is no longer user, rather the test_suite.clean_up.deletion() function is being
used to handle the issue of MS Windows not releasing the file in time.

Commented out some checks of the Relax_disp.test_hansen_cpmg_data_auto_analysis_numeric system test.

This is to allow this test to pass on 32-bit GNU/Linux systems. The numeric model optimisation is
incomplete but different between the 32-bit and 64-bit systems.

Loosened some of the relaxation dispersion system tests to allow them to pass on MS Windows.

Fixes to allow the Mf.test_mf_auto_analysis system test to pass on Mac OS X.

The simulated event.GetPosition() method in the Fake_right_click class the file
test_suite/gui_tests/model_free.py must return a wx.Point object and not a Python tuple. The
gui.components.base_list.Base_list.on_right_click() method has also been modified with a wx.Yield()
call to allos the test to pass.

Loosened the Relax_disp.test_sprangers_data_to_mq_ns_cpmg_2site system test to allow it to pass on Mac OS X.

Created a GUI element for the numeric_only flag of the auto-analysis for the dispersion GUI tab.

This defaults to false to allow all model types to be used.

Added a new button to the button bar in the relaxation dispersion GUI analysis tab.

This is a button used to launch the value.set user function to allow the user to pre-set certain
parameters so that they are not used in the grid search.

Completed the Relax_disp.test_hansen_cpmg_data_auto_analysis_numeric system test.

This now checks all the optimised parameter values and makes sure that no 'CR72' model was selected.

Implemented the numeric only option for the dispersion auto-analysis.

If the numeric_only flag is set to True, then no analytic models will be used in the final model
selection.