Test with other OpenAI RL algorithms

[moriartyjm]

Our current implementation uses the PPO algorithm. We should check that there is nothing special / particular about this choice.

• [x] Swap PPO for other standard OpenAI Gym RL solvers
• [x] Check learning curves and actions in both open and closed-loop modes