Bump version to 0.4.0

Modify use-env-states description in README (haosulab#39)

Simplify pick single reward to be more friendly to RL (haosulab#27)

* For PickSingle, change the original complex reward to simple reward, since the original reward can be unfriendly for RL, even though MPC can solve many objects through the original reward. The old reward is renamed as legacy.

Bump version to 0.2.0