[WIP] Load RL algorithm as generator in Xopt

[RemiLehe]

Merge #67 first

This adds a new notebook (RL_controller/Load_SB3_in_Xopt.ipynb) to try to load a stable-baseline algorithm into Xopt.

TODO:

• Fix remaining bugs that prevents the optimization from running
• Handle normalization of observations correctly, before passing observations to self.model.predict
• Handle de-normalization of actions after they are returned by self.model.predict and before they are passed to 1proposed_actuators = observed_actuators + action`
• Test that running the optimization indeed results in a low value of the reward function.