Bump SB3 version to ensure we have the bug-fix that makes the FloatReward unneeded.

Author: ernestum

setup.py

@@ -196,7 +196,7 @@ def get_local_version(version: "ScmVersion", time_format="%Y%m%d") -> str:
         "rich",
         "scikit-learn>=0.21.2",
         "seals~=0.2.1",
-        "stable-baselines3~=2.0",
+        "stable-baselines3~=2.2.1",
         "sacred>=0.8.4",
         "tensorboard>=1.14",
         "huggingface_sb3~=3.0",