Upload Chirayu and Shlok's MPE env wrappers and training script for MARL

[SimpleGeometry]

Draft PR. Putting this up for now so we can integrate this later. These environment wrappers (for MPE simple and tag) and training scripts were created by Chirayu Nimonkar and Shlok Shah, and I've created this PR on behalf of them with their permission.

Environments are in a good state, but need to integrate training script into the main JaxGCRL library (ideally we merge in clean_JaxGCRL in a way that's satisfying to the users). A direct integration with if statements and more flags is straightforward, but I hesitate to add complexity into the codebase (single-agent environment users will see more complexity from MARL which they won't use).

• We can of course just keep a separate MARL training file (either in cleanJaxGCRL or in main JaxGCRL) but this code duplication is prone to divergence when we need to update files.

For now, anyone that wants to use the MPE environments and train them can just use this branch and use the corresponding training file with cleanJaxGCRL.