Add one-command demo pipeline

[amarakramali]

Summary

• add a one-command synthetic demo pipeline
• generate sample data, features, two evaluations, plots, a trained model, and a future forecast
• keep every artifact under a configurable output root
• validate run settings before creating partial output
• allow lightweight estimator counts for development and integration tests
• make forecast plotting backend-independent unless interactive display is requested

Why

The no-download quickstart was reproducible but still required five manually coordinated commands and nine path arguments. A missed or mismatched path could leave a partial workflow. The new command applies one validated configuration to the complete pipeline.

Usage

python -m src.demo_pipeline

Use --output-dir to isolate all generated data, reports, figures, and models below another directory.

Validation

• python -m pytest -q — 65 passed
• uvx --from ruff==0.16.0 ruff check src tests streamlit_app.py — passed
• python -m compileall -q src tests streamlit_app.py — passed
• standard 90-day, 800-estimator command completed and generated all nine documented artifacts
• generated XGBoost evaluation reproduced 289.86 MW MAE versus 337.09 MW for the baseline