update RoI for the second L2MuonSA in L2 inside-out chains
This is a modification to fix the issue on L2 inside-out chains reported on ATR-22245. In L2 inside-out, L2MuonSA algorithm is called two times. The problem was that sometimes the first L2MuonSA gives negative decision to a RoI, but the second one gives positive decision. In order to fix the issue, RoIs already rejected in the first L2MuonSA are removed from the input of the second L2MuonSA.