the modification bring contunous CoM and ZMP along time but increase the error between the zmp ref and the zmp multibody

replacement of the RigidBodySystem class by the OnLineFootTrajectoryGeneration to compute the interpolation of the feet trajectory

Developping the herdt test to compute the difference between the multi body ZMP and the ZMP of reference
At this point the error is 5mm in X and 1mm in Y with average time consuption of 10ms per iteration

Adding security a security that block the computation of the dynamic filter while sending 0.0 velocity as command

	it creates files that can be used by openhrp to replay the solution
of the test on the robot in kinematics simulation mode

Creation d une nouvelle instance de CoMandFootRealizationByGeometrie pour le calcul du filtre dynamique dans Herdt.

try with continuous buffer, worst results... but maybe not comming from this calculation of the buffers.