fix: Update attack notebook to use OpenAI gpt-4o

[rdheekonda]

Summary

• Switch attacker model from Groq Llama 4 Maverick to openai/gpt-4o
• Read API keys from environment variables (CRUCIBLE_API_KEY, OPENAI_API_KEY)
• Fix VaultGuard GOAT cell using wrong transforms variable (transforms_toolshed → transforms_vaultguard)
• Add organization="dreadnode" to dn.configure
• Update setup markdown to reference OPENAI_API_KEY

Test plan

• Export CRUCIBLE_API_KEY and OPENAI_API_KEY env vars
• Run notebook end-to-end against prod
• Verify all 6 attack cells execute without errors

---

Generated Summary:

• Updated environment variable from GROQ_API_KEY to OPENAI_API_KEY to align with new API requirements.
• Changed the ATTACKER_MODEL from groq/meta-llama/llama-4-scout-17b-16e-instruct to openai/gpt-4o for improved model capabilities.
• Modified hooks parameter from transforms_toolshed to transforms_vaultguard, ensuring compatibility with new transformation processing.
• Overall adjustments enhance compliance with updated API standards and potentially improve performance and functionality of red teaming attacks.

This summary was generated with ❤️ by rigging