Added A2C models for Ant-v3, Humanoid-v3 and Walker2d-v3 environments

[pseudo-rnd-thoughts]

The SB3 benchmarks are missing the A2C Ant, Humanoid and Walker2d environments, benchmarks and zoo

This PR adds those environments for A2C

[araffin]

Hello,
thanks.
I think I will actually wait for Huggingface integration DLR-RM/rl-baselines3-zoo#198 (same as DLR-RM/rl-baselines3-zoo#244)

[pseudo-rnd-thoughts]

No worries

[araffin]

Basic version is ready to review ;)
DLR-RM/rl-baselines3-zoo#250

An example: https://huggingface.co/sb3/ppo-MountainCarContinuous-v0

[pseudo-rnd-thoughts]

Sorry, I didn't totally understand what you want me to do

[araffin]

Sorry, I didn't totally understand what you want me to do

sorry, we are going to replace the trained agents repo with huggingface hub, and the PR for that is ready, you can use it to upload models ;)

[araffin]

Closing in favor of DLR-RM/rl-baselines3-zoo#262
models are now uploaded on Hugginface hub, see https://huggingface.co/sb3/a2c-Walker2d-v3 for instance.
Btw, the pre-trained models do not have good performance and are missing some hyperparameter tuning (as expected, that was the reason I didn't upload them in the first place).