Add notebook DataFrame interface and examples

[gparadis]

Summary

• Adds modelwright.notebooks, a lazy pandas-backed notebook/DataFrame layer over ModelFacade.
• Adds helpers for declared inputs, outputs, scenarios, tables, reports, and baseline-vs-scenario comparisons.
• Adds the notebook optional extra with pandas>=2 while keeping pandas out of core runtime dependencies.
• Adds Sphinx notebook docs plus an Examples Gallery.
• Adds tracked examples for a tiny synthetic model and a production-size 2020 FABLE generated-model wrapper; the FABLE generated Python output is tracked as compressed .py.xz because the uncompressed module is above ordinary GitHub per-file limits.
• Bumps package/import/release metadata to 0.1.0a5.

Boundary

This release may claim initial notebook/DataFrame-facing helpers, scenario comparison workflows, and tracked synthetic/FABLE examples. It must not claim full spreadsheet UI behavior, automatic semantic recovery, stable public API compatibility, Excel-backed recalculation equivalence, or compact runtime IR production readiness.

Verification

• scripts/bootstrap_dev_env.sh
• .venv/bin/python -m ruff check .
• .venv/bin/python -m pytest -vv (163 passed, 1 skipped)
• .venv/bin/sphinx-build -b html docs _build/html -W
• .venv/bin/python scripts/verify_docs_theme.py _build/html
• MODELWRIGHT_RUN_FABLE_BENCHMARKS=1 .venv/bin/python -m pytest -vv tests/test_fable_wrapper_benchmark.py (1 passed in 149.67s)
• scripts/check_release_artifacts.sh (modelwright 0.1.0a5 clean wheel install and CLI smoke passed)

Closes #175.
Closes #178.
Closes #176.
Closes #177.
Closes #179.
Closes #181.
Refs #174.
Refs #180.