feat: add private local RAG quickstart with Ollama

[mhbuehler]

Summary

• Add a private local RAG quickstart using CrewAI Knowledge and Ollama.
• Use Sentence Transformers for local embeddings and ChromaDB for storage.
• Demonstrate task guardrail validation and automatic accelerator selection.
• Document Ollama as a model runtime option in the main README.

Validation

• Executed the notebook end to end with llama3.2:3b.
• Confirmed that execution produces a valid final response.