[GroupModel] (Re-)Calibration of the CentroidGroupModel

If we combine the CentroidGroupModel with PotentialFieldPedestrianCompactSoftshell and PotentialFieldObstacleCompactSoftshell the CentroidGroupModel produces unrealistic movement patterns in high density situations (see screenshot). In my opinion we have to increase "groupMemberRepulsionFactor" : 0.5 since the PedestrianPotential in case of CompactSoftshell is higher than for Compact. If one compares evacuation times there are lower if the group model is active which should not be the case!!

"groupMemberRepulsionFactor" : 0.01 (default)

"groupMemberRepulsionFactor" : 0.5