[RM-7129] spike rejection: try with smaller LSF size

Created on 2022-03-15 15:40:20 by Didier Vibert. % Done: 0

to reject solutions with unflagged hot pixels (or cosmics) one could test LSF sizes smaller than the given LSF size, when fitting for the best velocity.

one could for instance use a negative velocity dispersion and handle it in the linewidth computation (@CLineModelElement::GetLineWidth@) by reducing the profile size by : @- lambda*v^2/c@ instead of adding this term to the LSF size.

Note of course this will work only if the input LSF size is correct.

at the end: if a candidate corresponds to a negative velocity (or a velocity < to a given threshold) raise a warning or re-estimate all by flagging the concerned pixels (thus need to find those pixels, with the strongest line ? )