feat(forecast): add LightGBM feature-aware forecasting model (MLZOO-B)

[w7-mgfcode]

Implements PRP-30 (PRPs/PRP-30-lightgbm-first-advanced-model.md) — the first advanced feature-aware forecasting model.

Adds LightGBMForecaster (wrapping lightgbm.LGBMRegressor) on top of the leakage-safe shared feature-frame contract delivered by PRP-29 (MLZOO-A, #238).

Scope

• New LightGBMForecaster (requires_features=True), structural twin of RegressionForecaster.
• model_factory lightgbm branch (behind the existing forecast_enable_lightgbm flag).
• Scenario model_exogenous re-forecast dispatch generalised to requires_features.
• JobService._execute_train accepts model_type="lightgbm".
• lightgbm_version captured in the model bundle + registry runtime_info.
• LightGBM ships as an optional dependency group (ml-lightgbm); lazy-imported.

Out of scope

• No second model (XGBoost/Prophet → MLZOO-C).
• No feature-aware backtesting (deferred to PRP-MLZOO-B.2).
• No frontend, no explainability change, no migration, no API-contract change.

Follows PRP-29 (MLZOO-A, #238). Roadmap: PRPs/INITIAL/INITIAL-MLZOO-index.md.

[w7-mgfcode]

Verified shipped on dev. Fix commit: 2f1b8a5 feat(forecast): add LightGBM feature-aware forecasting model (#242) (PR #243). Code confirms: class LightGBMForecaster(BaseForecaster) at app/features/forecasting/models.py:625; ml-lightgbm = ["lightgbm>=4.5.0"] optional extra in pyproject.toml. MLZOO-B is complete.