relax SVN archive

Replaced all references to 'residue' by 'spin' in the relax_fit.mean_and_error() user fn docstring.

Set the default 'H_dim' variable in the intensity_xeasy() function.

Passing the variable to the intensity_xeasy() funciton the way it was done before could not have
worked...

Placed the setting of the 'H_dim' variable before the call to the intensity_xeasy() function.

The 'H_dim' variable was undefined otherwise...

This fix may break the handling of peak lists when the w1 and w2 dimensions are inverted.
Can these peak lists exist in XEasy ?

Modified the Modelfree4 system tests to work with the Portland C compiler Linux version.

The modelfree4 Linux binary compiled with Gnu gcc gives different results, sometimes quite large, to
the Linux binary compiled with the Portland C compiler. The tests now catch this. There may be
problems with the Mac OS X Modelfree4 binary still.

Grouped the three peak list support tests into a single one.

Continued the implementation of the XEasy support system test.

Much more work is needed, though.

Corrected the formatting of a few lines...

Made sure that the relaxation curve fitting time point is stored as a float.

The relaxation curve time point can now be an integer to allow the time of '0'.

See the post at https://mail.gna.org/public/relax-devel/2008-10/msg00040.html (Message-id:
<7f080ed10810141213s2211d36fre56906e0ae5903e2@...>).

Added a RelaxError to tell the user that the Modelfree4 version being used is buggy.

It also tells them that they must upgrade!