Corrected a rare instance of segmentation fault due to chi2 being rigourously 0

In the HybridSeeding createTracksFromHough, one has to remove multiple hits in the same layer, which is done by comparing chi2's of these hits wrt the track candidate. In some very rare instances, both hits are so close to the track candidate that the distance estimate is rigorously 0. In that case, the worstHit was never updated and pointed to anywhere, resulting in a bug.