Implement Multi-Agent Framework(s)

[jrcalgo]

Create an interface for users/developers to instantiate an async multi-agent training loop.

Frameworks needing implementation:

• CTCE (Centralized Training, Centralized Execution)
• CTDE (Centralized Training, Decentralized Execution)
• DTDE (Decentralized Training, Decentralized Execution)

The dynamics of MARL frameworks tend to function rather differently from one another, so there is much to take into consideration and likely much more to optimize for possible training scenarios.

I anticipate this functionality's performance will benefit greatly from a rewrite in a lower-level language like Rust or C++ (to keep it in scope for now, it will be written in Python).

[jrcalgo]

This will be implemented in Rust once the single-agent framework is rewritten and merged into main.

Rust implementation will provide Python bindings just like the single-agent framework