Bump verifiers

[mikasenghaas]

Bump verifiers to latest main, notable changes:

• #657: Improvements to math rubric
• #658: Fix import for PythonEnv

Also removes prime-sandboxes from pyproject, as its an environment dependency.

Also adds configs that plays around with updated INTELLECT-3 training environments from #48 in research-environments.

---

Note

Adds env-mix RL configs (incl. math-python sandbox) and bumps verifiers to cdbc417 while removing prime-sandboxes from project deps.

• Configs:
  • Add configs/env_mix/env_mix.toml and configs/env_mix/README.md for mixed-environment RL.
  • Orchestrates math, code, science, logic, and math-python (with sandbox, python tool, and resource limits).
  • Sets training/inference params (GPU IDs, batch size, rollouts, WandB, Qwen/Qwen3-4B-Instruct-2507, tool-call parser, max model len).
• Dependencies:
  • Update verifiers source rev to cdbc417 in pyproject.toml.
  • Remove prime-sandboxes from top-level project dependencies.

^{Written by Cursor Bugbot for commit 7b07785. This will update automatically on new commits. Configure here.}

[samsja]

lgtm