Add RL on Verifiable Reward (RLVR) reference implementations

[jacobthebanana]

This pull request includes:

• Minimalized TRL/VeRL to run one PPO optimization step, given token array, per-token advantage array, model checkpoint, and PPO hyperparameters.
• Agent SDK integration- define the environment using the familiar OpenAI Agent SDK and run RL on the LLM powering the agent. Not yet tested on multi-agent setups (agent as tool or handoff)
• Extensive typing for simplified function signatures and IDE support- static type checking, pyright lints, proper autocompletion even within the training loop.

Vec-Inf wrapper integration will come in a separate pull request.