Add a model-vs-baseline benchmark script

CHANGELOG.md

@@ -6,6 +6,7 @@ Format: [Keep a Changelog](https://keepachangelog.com/en/1.1.0/)
 ## [Unreleased]
 
 ### Added
+- `scripts/benchmark_models_vs_baselines.py`: pairs every estimator with the simple domain rule it is meant to replace (rank by last year's total, predict last gift, mail everyone) and evaluates both on held-out, walk-forward splits: top-1%/5%/10% hit rate and lift, ROC-AUC, average precision and decile calibration for classifiers; MAE and within-25% for amount predictions; net revenue/ROI for cost-aware mail selection. Runs on the synthetic donor panel (five seeds, mean and min-max) by default; `--skip-kdd98` stays fully offline, `--fast` gives a one-seed smoke run, and `--out` writes the results table to JSON and CSV. Separate from the existing `scripts/benchmark_models.py` per-model accuracy table, which this does not replace or touch.
 - `philanthropy.ingest.map_columns(df, mapping, *, required=...)`: renames a
   user-supplied export's headers to canonical names and raises one
   `ValueError` listing every still-missing required column, for callers

docs/explanation/benchmarks.md

@@ -59,6 +59,13 @@ script:
 python scripts/benchmark_models.py
 ```
 
+Each model's accuracy above is measured alone. For whether a model actually
+earns its complexity over the simple rule it would replace (rank by last
+year's total, predict last gift, mail everyone), see
+`scripts/benchmark_models_vs_baselines.py`, a separate script pairing every
+estimator with its naive baseline on held-out, walk-forward splits of both
+the synthetic donor panel and KDD Cup 1998.
+
 For each of **five seeds** (42–46) the script builds a 4,000-row synthetic pool,
 takes a stratified 75/25 train/test split, and fits every applicable binary
 classifier in `philanthropy.models` on the documented feature set