Model based reinforcement learning

[findmyway]

After taking a look into facebookresearch/mbrl-lib . I think our current design is flexible enough to implement most algorithms in it. It's just that we need to standardize some interfaces.

• A SURVEY ON MODEL-BASED REINFORCEMENT LEARNING

[albheim]

Just mentioning that I have started looking at this a little. Not promising anything or saying that anyone else shouldn't try to make it, but might have a PR for at least one of the algorithms soon and I'm interested in the other stuff too so might have a go when there is time.

I had some question about organization of the stuff, but I think we can tackle that in the PR when I have also thought through it a bit more.

EDIT: See #531