Update and unify SciFi cluster resolutions
The SciFi hit resolutions we use in the HLT2 pattern reco are off by a factor of ~2 (i.e. the PrKalmanFilter simply uses them divided by 2 see https://gitlab.cern.ch/lhcb/Rec/-/blob/b97ed24bae61eeaeb778b212c72e58c66241544f/Tr/PrKalmanFilter/include/PrKalmanFilter/KF.h#L64). While this is not super dangerous, it is definitely suboptimal as it causes the KalmanFilter to easily consider hits as outliers which were fine in the pattern reco only because of this discrepancy, and might be one reason why we observe relatively many ghosts at high momenta where track uncertainties are dominated by hit resolution.
Related to #110 and !1941 (closed)