Small differences in matching track momentum after fixing sign

Fixing the sign of qop in Seeding&Matching long tracks (!1210 (merged)) introduced some small differences in the absolute momentum. The effect is less than 1% when looking at the validators, and so far seems to be coming from the small correction https://gitlab.cern.ch/lhcb/Allen/-/blob/master/device/track_matching/match/src/TrackMatchingVELOSciFi.cu#L112 (factor * dev_magnet_parametrization->momentumParams[7]).

Current plan: look at momentum residuals and cross-check with Rec equivalent to make sure the correct sign convention is used.

@rmatev @cmarinbe @mstahl