computation results.
- bug #2102708: Weighted linear and polynomial fits required actual weights
(1/sigma^2) to be specified in place of standard errors, while all others
(as well as error bar plots) needed the standard errors themselves.
Solution: Switched everything to using standard errors, eliminating the term
"weighting" from the GUI and most of the source code.
Rationale: Avoid confusion for users, especially given the previous
inconsistent treatment of the matter. Avoid confusion for
devs. Arbitrary weights don't work naturally with non-linear fits;
and they almost certainly lead to nonsensical error estimates of
results (unless you _really_ know what you're doing, in which case
you can simply specify 1/sqrt(weight) as standard error).
- Fixed estimation of parameter errors in case of (scaled) Levenberg-Marquardt
fitting with unknown Y errors. See
http://www.gnu.org/software/gsl/manual/html_node/Computing-the-covariance-matrix-of-best-fit-parameters.html
- Fixed computation of R^2 in case of known Y errors.
