Change a,r,s' interface to s,a,r,s'

[kengz]

Change interface to s,a,r,s'

• change all the interface from action, reward, state to state, action, reward, next_state and propagate in all methods
• this is a safe interface change without affecting logic. all changes are traceable and complete. Procedure: add new argument, rename old state to next_state.
• this means the internal variable self.last_state of memory is no longer needed
• changes done in:
  • control loop for singleton and space, which affects
  • agent.update, which affects
  • memory.update, which affects
  • memory.base_update and memory.add_experience