feat(forecast): add Prophet-like additive forecasting model (MLZOO-C2)

[w7-mgfcode]

Implements PRPs/PRP-MLZOO-C2-prophet-like-additive-model.md.

Add ProphetLikeForecaster — a distinct model-family design, not a tree-model clone and not the real prophet package:

• Pure scikit-learn: Pipeline([SimpleImputer(median), Ridge(solver="cholesky")]) over the canonical 14-column feature frame.
• Additive structure with a decompose() method (trend / seasonality / regressor contributions).
• No optional dependency, no feature flag, no metadata change — structurally like the existing regression baseline.
• Works in train / predict / scenarios / backtesting via the capability-based requires_features contract (B.2) with no service-layer edits.

Independent of and parallel to MLZOO-C1 (XGBoost). See PRPs/INITIAL/INITIAL-MLZOO-C-xgboost-prophet-extensions.md and INITIAL-MLZOO-index.md (roadmap row 3b).

[w7-mgfcode]

Merged to dev via #250 — ProphetLikeForecaster implemented end-to-end, all CI gates green. Will reach main on the next release.