Test impact of outliers in lightcurves and remove them if necessary

[EiffL]

@hainow As you may see if you plot a few lightcurves, there are some outliers that remain despite my best efforts to correctly match objects between runs. So there are two questions here, 1) do they matter ? i.e. do they impact the classification accuracy 2) Can we clean them out ?

If you have some time to look into this, I am particularly curious about the first question. Meanwhile I can try to find out why these outliers are there and remove them.