Fix extrapolation to beamspot when calculating track ClosestToBeam states

Take into account beamspot position (instead of (0,0)) when extrapolating tracks to their ClosestToBeam state (as was done for Allen in Allen!1085 (merged))

Enclosed, a comparison between the track.state() in Allen and the track.closestToBeamState() coordinates in Rec w and w/o applying this fix, obtained by running on the same event in a 2022 run. The large x/y/z differences are removed.

An HLT1-HLT2 states comparison test will follow in a separate Allen+Moore MR (similar tests are atm blocked by prerequisite MRs, so I thought was better to fix this in the meanwhile).

cc @cmarinbe @dovombru @ahennequ