feat: support separate red and blue model endpoints

[Ker102]

Goal

Allow nullstate to use separate OpenAI-compatible endpoints for red-team and blue-team agents, so hackathon demos can run two vLLM containers on different local tunnel ports.

Acceptance Criteria

• NULLSTATE_RED_LLM_BASE_URL and NULLSTATE_BLUE_LLM_BASE_URL override the shared NULLSTATE_LLM_BASE_URL per role.
• --red-base-url and --blue-base-url CLI flags are available for explicit runs.
• Metrics artifacts show role-specific endpoint collection when the endpoints differ.
• Existing single-endpoint behavior remains compatible.
• Docs explain one-container and two-container usage.

Notes

Stack this on PR #13 and leave the PR open for evidence capture.

[coderabbitai]

🔗 Related PRs

#7 - feat: collect model serving metrics artifacts [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!