40 finish first version of gym interface

[Armandpl]

Opening PR as draft because there is at least one question that needs answering before finishing this.

• do we want to implement a command to reset only the motor encoder? -> no
  • here is how we handle env reset now: we wait for the pendulum to stay in a cone for a set amount of time after which we consider it is at rest and we reset BOTH the motor encoder and the pendulum encoder. This is making me think there is a small chance the pendulum might still be moving at that time or maybe it stopped slightly off from 0 because of frictions. This means when resetting the pendulum the zero is going to move a little bit.
    • is this an issue?
    • even if we add the option not to reset the pendulum, the zero is set when we power on the robot so we have the same issue? if the pendulum is slightly off the zero is going to be slightly off? -> so no need to implement this

[pierfabre]

see comments, apart from that lgtm