Make app reproducible (datasets + E2E run)

[pypi-ahmad]

This PR aims to make the project easier to run/test locally and more contribution-ready, while keeping the Streamlit app functional even when market-data providers are rate-limited.

Changes

• Streamlit: added a Price History Source selector with fallback options (yfinance, local_panel, plotly, and auto) so the app remains usable if Yahoo Finance returns HTTP 429.
• UX: clearer CSV validation for Batch Forecast vs Fine-Tune panel datasets.
• Tooling: added scripts/download_test_datasets.py to generate 3 small datasets under data/samples/ for manual testing.
• Tooling: added scripts/e2e_real_run.py to run a real end-to-end flow (TimesFM inference + optional Ollama report + PDF export) and write artifacts to /tmp.
• Docs: improved README and added CONTRIBUTING.md, SECURITY.md, and a PR template.
• Community: added MIT LICENSE and CODE_OF_CONDUCT.md.

Manual Verification

• pytest -q (51 tests)
• ./.venv/bin/python scripts/download_test_datasets.py
• ./.venv/bin/python scripts/e2e_real_run.py --data-source plotly (with Ollama running locally)
• streamlit run app.py

Notes

• data/ is gitignored; the sample datasets are generated by the script instead of being committed.