Fix eta calculation for backward VELO tracks

While commissioning, it was noticed the eta calculation for VELO tracks, based on the momentum, gave nan. With this MR, the state() method on VELO tracks returns a KalmanVeloState instead of a KalmanState (using the conversion already available) and a method for eta calculation based on the slopes is added. Nothing changes for eta calculation of Basic/CompositeParticle.

By monitoring VELO tracks eta on simulation, the obtained distribution is shown below

Also, Velo::Track and Velo::TrackFitParameters structs, unused, are cleaned