evaluate recurrent RL algos

[Armandpl]

• research if there exist other algorithms apart from recurent PPO?
• in the gym wrapper, add the option to mask out the speed
  • think about safety? do we continue filtering position to get speed?
  • i think we should just try a run without speed limits and see if it looks it could break the robot?
  • i think we should continue computing the speed unless it takes too much time AND we want to increase the control freq. see write script to bench the max control freq when using PPO, SAC etc #55