Train set of parameters for more stability with reduced walking speed

[texhnolyze]

As we do not have the same walking speed in reality as in simulation and the parameter optimization is using walking speed as its reward optimization, we want to try to improve stability while capping walking speed to realistic speeds.