PrKalmanFilter: fix creation of Rich1 states

This should fix the problems we had with the PrKalmanFilter's RICH1 states, reported in Panoptes!116 (merged).

@jonrob Thanks for creating that test in Panoptes 👍 Here are some plots from that test:

Default test options using TrackMasterFitter
PrKalmanFilter, and no TrackSelection

Goes together with a new test in Panoptes!120 (merged) and refs in Moore!808 (merged)
cc @ldufour @wouter