[OSPP-Week4] Debug and write PLAS algorithm

[pilgrimygy]

@findmyway
Last week plan

Last week, I did the work:

I will try to complete the continuous version of the Critic Regularized Regression (CRR) algorithm. But its performance is poor.

Week 4:

1. I will fix the bugs in Continuous CRR.
2. Begin to write the PLAS algorithm.
3. Consider adding a new component (for example, VAE) into RLCore.

[findmyway]

Cool, let me know if you need any help in debugging the performance issue.

[pilgrimygy]

#407

Maybe the actor_loss_coef calculation is wrong. If it doesn't use actor_loss_coef to weight, the algorithm will be similar to Behavior Cloning and its performance will be slightly better.