Understand/Correct momentum bias of Long Tracks (fitted without UT hits)

When fitting Long Tracks in the "without_UT" scenario the TrackResChecker shows a biased momentum pull: hlt2_light_reco_pr_kf_without_UT_with_mcchecking on x86_64_v3-centos7-gcc11-opt+g lhcb-master/1752 (29.7.2022)

TrackResCheckerBestLong                INFO Long/x pull     :  mean =  0.009 +/- 0.015, RMS = 1.060 +/- 0.013
TrackResCheckerBestLong                INFO Long/y pull     :  mean =  0.010 +/- 0.015, RMS = 1.090 +/- 0.015
TrackResCheckerBestLong                INFO Long/tx pull    :  mean =  -0.005 +/- 0.015, RMS = 1.043 +/- 0.013
TrackResCheckerBestLong                INFO Long/ty pull    :  mean =  -0.005 +/- 0.015, RMS = 1.086 +/- 0.015
TrackResCheckerBestLong                INFO Long/p pull     :  mean =  -0.623 +/- 0.019, RMS = 1.308 +/- 0.017

This is not (so much) the case including UT hits: hlt2_light_reco_pr_kf_with_mcchecking on x86_64_v3-centos7-gcc11-opt+g lhcb-master/1752 (29.7.2022)

TrackResCheckerBestLong                INFO ALL/x pull     :  mean =  0.010 +/- 0.015, RMS = 1.075 +/- 0.013
TrackResCheckerBestLong                INFO ALL/y pull     :  mean =  0.004 +/- 0.016, RMS = 1.120 +/- 0.015
TrackResCheckerBestLong                INFO ALL/tx pull    :  mean =  -0.006 +/- 0.015, RMS = 1.070 +/- 0.014
TrackResCheckerBestLong                INFO ALL/ty pull    :  mean =  -0.001 +/- 0.016, RMS = 1.113 +/- 0.015
TrackResCheckerBestLong                INFO ALL/p pull     :  mean =  -0.100 +/- 0.019, RMS = 1.304 +/- 0.018

This might be the reason for the sightly shifted D0 mass peak as shown in Reco_Comparison__1_.pdf. Behaviour of TMF without UT has not been checked AFAIK.

/cc @chasse @decianm @ausachov @wouter @pherrero