Use a Kalman state with CompassUT

With !410 (merged), several Kalman states are introduced: One at the beamline, another at the end of the VELO. Both these states were meant to supersede the previous least means square fit beamline state, since the new states have better precision. However, CompassUT behaves better with the lms fit beamline state when compared to any Kalman state.

This should be studied and understood. Most probably it is possible to improve the performance of CompassUT by using a Kalman state and changing the logic of the algorithm. When this is done, the lms fit beamline state can be removed from algorithm VeloKalmanFilter.

Merge request: !514 (merged)