Upgraded ray==1.10.0, tensorflow==2.8.0, stable_baselines3 example, tune/train/validate on different datasets in examples/use_attentionnet_rllib.ipynb, added finta, tuneta, s3_contrib, tensorboard output in Docker, PPO example on stable_baselines with rendering, fixed fees always over zero and automatic rounding to precision if more than 0, so no crash... #396