[codex] Bundle cookbook example data and remove yfinance

[whenpoem]

Summary

• add a packaged pypfopt.data module with loaders for bundled example prices, SPY prices, and Black-Litterman market caps
• update the 5 cookbook notebooks to load bundled example data instead of downloading with yfinance
• remove yfinance from the dev extra and add tests for the new data loaders

Why

Issue #716 tracks the cookbook notebooks' dependency on live yfinance downloads. That makes notebook CI depend on external network calls and changing upstream data. Bundling the example datasets keeps the notebooks reproducible and allows nbmake to run offline.

Impact

• cookbook notebooks now run against stable packaged example data
• package distributions include the example CSV and JSON assets used by the notebooks
• developers no longer need yfinance installed to run the notebook test suite

Validation

• uv run --python .venv\Scripts\python.exe pytest ./tests
• uv run --python .venv\Scripts\python.exe pytest --reruns 3 --nbmake --nbmake-timeout=3600 -vv cookbook
• uv run --python .venv\Scripts\python.exe python -m build --sdist --wheel --outdir dist-check

[atharvajoshi01]

Makes sense to bundle the data, same thing sklearn does with load_iris etc. Notebooks should just work now without needing a live yfinance connection. One thing I'd double check is whether the bundled CSVs match what the original yfinance calls were returning, just so the notebook outputs stay consistent.

[whenpoem]

Makes sense to bundle the data, same thing sklearn does with load_iris etc. Notebooks should just work now without needing a live yfinance connection. One thing I'd double check is whether the bundled CSVs match what the original yfinance calls were returning, just so the notebook outputs stay consistent.

Thanks for the careful check.the first version bundled data that made the notebooks run offline, but it did not fully preserve the original yfinance ticker universes and outputs.
so I’ve pushed a follow-up commit that:
restores the original notebook ticker lists, constraints, and Black-Litterman views;
adds bundled cookbook price data matching the original yfinance download calls;
updates the SPY data and market-cap snapshot used by the Black-Litterman notebook;
adds tests to ensure the bundled data covers all tickers required by the notebooks.
I also re-ran the notebook suite and package checks:
pytest --nbmake --nbmake-timeout=3600 -q cookbook: 5 passed
pytest tests -q: 316 passed, 5 skipped
python -m build --sdist --wheel: successful
I additionally spot-checked the new bundled CSVs against fresh yfinance downloads with the same ticker/date contract, allowing only tiny floating-point/CSV roundtrip differences.(the maximum relative difference was around 2e-6)

[whenpoem]

Makes sense to bundle the data, same thing sklearn does with load_iris etc. Notebooks should just work now without needing a live yfinance connection. One thing I'd double check is whether the bundled CSVs match what the original yfinance calls were returning, just so the notebook outputs stay consistent.

Hi @atharvajoshi01, thanks again for the review. I pushed a follow-up commit addressing the bundled CSV consistency issue you mentioned.

Could you please take another look when you have a chance and let me know if this resolves your concern?

Also, the PR still seems blocked by pending workflow approvals / required review. If someone with write access could approve the workflows, that would be very helpful. Thanks!