Jacklanchantin/verifiable nonverifiable

[jacklanchantin]

What does this PR do? Please describe:
A summary of the change or the issue that is fixed.

Fixes #{issue number}

right now, we pass all prompts to the LLM reward model for synchronization. need to optimize it. e.g.:

in SuperReward:

1. collect all prompts and task types on rank 0
2. depending on task type submit all prompts to their rewards
3. carefully scatter that back to their ranks

the biggest mismatch is that non-LLM rewards all run in parallel, but with LLM ones we have to do collective call, so here we have to unify that

Does your PR introduce any breaking changes? If yes, please list them:
List of all backwards-incompatible changes.

Check list:

• Was the content of this PR discussed and approved via a GitHub issue? (no need for typos or documentation improvements)
• Did you read the contributor guideline?
• Did you make sure that your PR does only one thing instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests?
• Did you verify new and existing tests pass locally with your changes?
• Did you update the CHANGELOG? (no need for typos, documentation, or minor internal changes)

[jacklanchantin]

if i check for if self.vllm_math_reward_model is not None which it will be None for non combination runs, then i get an nccl error. not sure what's happening there