Added ProximalPolicyOptimizationClip model.
Added ReinforcementLearningActorCriticNeuralNetworkBaseModel model.
Refactored the codes for ProximalPolicyOptimization, VanillaPolicyGradient, ActorCritic and AdvantageActorCritic so that it inherits from ReinforcementLearningActorCriticNeuralNetworkBaseModel.
Episode updates now runs at the end of final reinforcement at the same episode. Previously it runs before the first reinforcement at the next episode.