fix sign of polarity in seeding and matching

Fix sign of tracks reconstructed from seeding & matching. A wrong sign assignment was introduced with the change of parametrization in !1098 (merged) . The sign reversal was universal, so at the time rates of lines with OS requirements were unaffected. The issue was observed from @mstahl when running hlt1_pp_forward_then_matching sequence where Lambdas were mostly reconstructed as same-sign (probably with one "forward" proton and one "matching" pion with the wrong sign).