Both simulation and correction of Lorentz shift in SPD was not accounting for
1) collected holes move against sensor local Y axis
2) SPD2 has its Z axis opposite to lab Z axis, hence in local frame the Bz is inverted
As a result, both simulated and corrected Lorentz shift on SPD1 had its sign flipped. The effect
in MC is 0 (since error was in both), for raw data the effect should be ~ -/+7 micron bias of
<DCA> @ vertex for B+/B- respectively, leading to ~15 micron split of <DCA> bias between
2 polarities