[metalearn] support PPO training algorithm

[cosmicBboy]

To improve stability and robustness of policy, implement proximal policy optimization (PPO):

• https://arxiv.org/abs/1707.06347
• code: https://github.com/ikostrikov/pytorch-a2c-ppo-acktr-gail

[cosmicBboy]

https://github.com/ikostrikov/pytorch-a2c-ppo-acktr-gail/blob/master/a2c_ppo_acktr/algo/ppo.py#L61-L66