[Bug-gsl] parameters error nonlinear fit

[Massimo Pica Ciamarra]

I guess there is a bug-imprecision in the
example relative non linearfit least square fit.

In order to get an estimate of the parameter errors,
when the reduced chi square is substantially different from 1,
modify the code as follows:

double ft = sqrt(gsl_blas_dnrm2(s->f)/(n-p));
#define FIT(i) gsl_vector_get(s->x, i)
#define ERR(i) sqrt(gsl_matrix_get(covar,i,i))*ft