Experimental launch script

[ffelten]

• Make a CLI (cleanRL-like) to launch experiments on different envs;
• Add autotag feature for reproducibility;
• Add multi policy logging to MOMAPPO;
• Script to check the resulting policies on rendered env;
• Fix versions properly (compat with Python 3.8);
• Run MOMAPPO on MultiWalker and CrazyRL;
• Test on HPC

Example command line to launch:
python momaland/learning/continuous/cooperative_momappo.py --env-id surround_v0 --seed 0 --ref-point -10 -10 --track

Generates this kind of beauty: https://wandb.ai/openrlbenchmark/MOMAland/runs/dsfo5vyz/overview