relax SVN archive

To get the distribution, one should look at the sum of squares.
But this should be from the weighted sum of squares, which is stored as chi2.
The API provides a method of model_statistics(), to get:
k - number of parameters.
n - number of data points.
chi2 - the chi-squared value.

Then reduced chi2 can be computed. chi2_red = chi2/(n-k)
And the error distribution be drawn from this.

The command used was:
svn merge -r27213:r27201 .

Task #7882 (https://gna.org/task/?7882): Implement Monte-Carlo simulation, where errors are generated with width of standard deviation or residuals.