refactored RL code, setting seed inside sampler function, evaluate can be...

refactored RL code, setting seed inside sampler function, evaluate can be constrained to only output specific goals, rule based simulator samples max_initiative