Update env reset interface

[kengz]

Update env reset interface

• update env.reset interface to be consistent with gym return: state. all logic perserved
• methods involved:
  • env.reset
  • env.space_reset
  • env.space_step
  • env_space.reset
• purge memory.base_update for simplicity and update memory.update to remain equivalent

Memory/Misc

• move total_reward from memory into body, improve calculation and logging
• remove onpolicy memory warn_size_once to simplify code since sample already checks for flushing
• rename memory.true_size to memory.size
• add reward_ma to body.df
• purge debug2, debug3 logger methods