relax SVN archive

A few fixes for the model-free duplicate() method.

Removed the creation of the model selection data pipe, as this is now automatic.

This commit follows r6625 where the same change was applied to the full_analysis.py sample script.

Rearranged the pipe switching before data duplication.

Changed the pipe switching behaviour of the model_selection function.

Rather than finishing on the original current data pipe, as the function creates a model selection
pipe then the function finishes by switching to that pipe.

Added some pipe switching to the model selection code.

The current pipe at the start is stored, and then restored at the end.

Removed the attempt at sending in pipe_from into the determine_param_set_type() method.

Started to convert the model-free duplicate_data() method to the new relax design.

Removed the creation of the model selection data pipe, as this is now automatic.

Removed the check for the same sequence data in all pipes.

For flexibility, this check should not occur as some models are sequence independent. And the
checks weren't functional anyway.

Import fixes for the module.