[RM-6508] adaptative 2nd pass redshift window size

Created on 2021-05-21 08:24:00 by Didier Vibert. % Done: 50

given the behavior of the distance shift from 1st pass candidate to 2nd pass solution (see #6051, and attached plots: https://projets.lam.fr/attachments/7362 & https://projets.lam.fr/attachments/7363),

if we want to be conservative to keep all candidates future solution inside the second pass window, we need a big enough size, which will waste some time for the majority of spectra, thus an alternative is to handle specifically the small number of spectra for which the solutions moves away.

Morevover , from #6360, at least some of them are explained by large velocity dispersion leading to a larger pdf peak width.
Several possibility (not exclusive):

• if no maxima is found in the second pass window (max is at border) enlarge the window size (only on the size where the max is at border ? ) and recompute 2nd pass
• since 2nd pass window is used after fitting the velocity dispersion, one could redefine the window size given the fitted velocity, for large velocity we could enlarge the window size before even starting to use the 2nd pass window (@COperatorLineModel::RecomputeAroundCandidates@)