Morphological Computation + Soft Actor-Critic

Working on a version of SAC with Morphological Computation as intrinsic reward to boost the performance or regularize an over-actuated system. One very related paper: https://openreview.net/forum?id=DXJl9826dm

• Implement MC with cleansac
• Add hyperopt scripts for Ant and FetchEnvs
• Render i_reward, e_reward, mc_mean and mc_std
• Fix logging of rewards (training and eval)