Environment Seeds
We should implement seeding, so that we can pass a custom seed into a env.seed()
call, as referenced in the interface here : https://github.com/openai/gym/blob/master/gym/core.py#L115
This is critical to ensure that after stochasticity is introduced, all the evaluations are comparable (as they will use the same set of seeds), and also to ensure that the local instance of the env doesnt diverge from the remote instance of the env, during evaluation.