feat: Open-Meteo catchment rainfall series + rain features + harness variant

src/ml/rain.py fetches hourly precipitation for five upper-Ping
catchment points (Chiang Dao, Mae Taeng, Mae Ngat, Mae Rim, city) from
the Open-Meteo forecast-model archive (2021-03 onward, no API key,
Bangkok-local timestamps, year-chunked local cache) plus the live
forecast endpoint for serving (trailing days + next 48h).

features.build_features/build_matrix accept the catchment-mean series
and add rain_6h/24h/72h trailing sums and rain_fc24 — the forward 24h
sum, a genuine forecast feature (archived forecasts at training time, a
real weather forecast at serving; never contains river data). Columns
exist only when a series is provided: HGB rejects all-NaN columns at
fit, so no-rain training omits them and serving passes an empty series
for alignment. evaluate.py gains the rise_rain variant (and --no-rain)
so the harness can judge whether rain beats the deployed rise baseline.

src/ml/features.py

@@ -200,8 +200,24 @@ def _hours_since_observed(mask_col: pd.Series) -> pd.Series:
 # ---------------------------------------------------------------------------
 
 
-def build_features(grid: HourlyGrid, station: str) -> pd.DataFrame:
-    """Build the deterministic-order feature matrix for one target station."""
+RAIN_FEATURES = ("rain_6h", "rain_24h", "rain_72h", "rain_fc24")
+
+
+def build_features(
+    grid: HourlyGrid, station: str, rain: Optional[pd.Series] = None
+) -> pd.DataFrame:
+    """Build the deterministic-order feature matrix for one target station.
+
+    ``rain`` is the hourly catchment-average precipitation series (Open-Meteo,
+    src/ml/rain.py). Rain columns are added only when a series is passed:
+    HistGradientBoosting REJECTS all-NaN columns at fit time, so training
+    without rain must omit the columns entirely (bundles record their
+    feature_names, and serving subsets to them). At serving, pass an empty
+    series rather than None so the columns exist (as NaN) for rain-trained
+    bundles even when the live fetch fails. rain_fc24 is the forward 24 h
+    sum: the archived forecast series at training time, a real weather
+    forecast at serving time; it never contains river data.
+    """
     idx = grid.observed.index
     cols: Dict[str, pd.Series] = {}
 
@@ -254,6 +270,16 @@ def build_features(grid: HourlyGrid, station: str) -> pd.DataFrame:
     cols["doy_cos"] = np.cos(2 * np.pi * doy / 365.25)
     cols["is_monsoon"] = idx.to_series().dt.month.isin(MONSOON_MONTHS).astype(float)
 
+    if rain is not None:
+        r = rain.reindex(idx)
+        cols["rain_6h"] = r.rolling(6, min_periods=1).sum()
+        cols["rain_24h"] = r.rolling(24, min_periods=1).sum()
+        cols["rain_72h"] = r.rolling(72, min_periods=1).sum()
+        # forward sum over (t, t+24]: shift(-1) starts the window at t+1
+        cols["rain_fc24"] = (
+            r.shift(-1).iloc[::-1].rolling(24, min_periods=1).sum().iloc[::-1]
+        )
+
     return pd.DataFrame(cols, index=idx)
 
 
@@ -331,10 +357,11 @@ def build_matrix(
     station: str,
     horizons: Tuple[int, ...] = (6, 12, 24),
     stats_end: Optional[str] = None,
+    rain: Optional[pd.Series] = None,
 ) -> Tuple[pd.DataFrame, pd.DataFrame, dict]:
     """Build (X, Y, meta) training/inference matrices for one station."""
     grid = make_hourly_grid(df_long)
-    X = build_features(grid, station)
+    X = build_features(grid, station, rain=rain)
     Y = build_labels(grid, station, horizons, stats_end=stats_end)
 
     keep = X["obs_age_h"].notna()