Postfit parameters should be user settable and recorded

Created by: Zaharid

The chi² and arclength filters in postfit are probably the worst part of the methodology: We have some parameters that control heavily the final result which have never been properly investigated and don't come from anywhere: We are assuming for no particular that the distribution of chi² and arclength should be gaussian and we discard things that are outliers given 1000 replicas (no matte how many you have; in reality the parameters should scale with sqrt(nrep) but..) and the former assumption.

The idea situation would be getting rid of these filters altogether, but firstly we should at least be able to control the parameters and record the settings that are actually being used.