Add recipe recommender domain demo

[maxrross]

Summary

• Adds a recipe and meal-planning recommender domain with a local 12-recipe catalog.
• Encodes recipe attributes plus meal context into binary AO inputs: cuisine, meal slot, cook-time bucket, difficulty, dietary flags, requested meal slot, max cook-time bucket, dietary preference, and mood.
• Adds a Streamlit recipe demo that uses a live AO Agent when ao_core/ao_arch are available and a deterministic fallback scorer when private AO packages are not installed.
• Adds a matching AO architecture, README instructions, and focused unit tests for encoding, scoring, and recommendation behavior.

Bounty Context

For aolabsai/ao_pyth#9.

Validation

• python3 -m unittest discover -s tests
• python3 -m py_compile recipe_domain.py recipe_recommender.py arch__RecipeRecommender.py tests/test_recipe_domain.py
• git diff --check
• Temporary Streamlit smoke in a clean venv: streamlit run recipe_recommender.py served HTTP/1.1 200 OK on 127.0.0.1:8562; direct script execution completed without AO packages, exercising the fallback path.

Transparency

AI-assisted with Codex. I reviewed the diff and ran the validation above before submitting.