cookbook(tracing): add Chroma RAG evaluation notebook

[SuhaniNagpal7]

Summary

Companion notebook for the docs cookbook "Trace and Evaluate a Chroma RAG Pipeline with Future AGI" (future-agi/docs#TBD). Mirrors the docs walkthrough as a runnable `.ipynb`:

1. Auto-instrument ChromaDB and OpenAI with `traceAI`.
2. Build an in-memory Chroma collection with five short policy snippets.
3. Run a retrieve-then-generate flow on `gpt-4o-mini`.
4. Score the result with five Future AGI evaluators (`context_relevance`, `chunk_attribution`, `chunk_utilization`, `completeness`, `factual_accuracy`).

Includes interpretation, troubleshooting, and next-steps sections so the notebook stands on its own when opened from a Colab badge.

Branch convention

Following the pattern used by other docs-companion notebooks (e.g. `cookbook/quickstart-notebooks`), this notebook lives on its dedicated cookbook branch. The docs Colab badge will point at `/blob/cookbook/evaluate-chroma-rag/tracing/evaluate-chroma-rag.ipynb`.

Verification

The exact Python in this notebook was run end-to-end against a live Future AGI workspace before being committed. `register()` connected, `ChromaDBInstrumentor` + `OpenAIInstrumentor` instrumented, `agent` answered, and five eval scores landed (Context Relevance 1.0, Chunk Attribution 1.0, Chunk Utilization 0.6, Completeness 1.0, Factual Accuracy 1.0). Three spans landed in the FAGI dashboard.

Test plan

• Open the Colab badge from the docs PR once both PRs are merged.
• Run all cells top to bottom with `FI_API_KEY`, `FI_SECRET_KEY`, and `OPENAI_API_KEY` set; confirm one Chroma trace and one OpenAI trace land in the dashboard, plus the five-row score printout.