Fix: Use issued_at instead of target_at for zero-hour resolved truth timestamps

The zero-hour forecasts (observations) were being saved with resolved_at = target_at,
which could be in the future. This caused them to be filtered out by the time-based
query in load_display_dataset(). Now using issued_at (when the observation was made)
ensures the timestamps fall within the valid display range.

debug_weather.py

@@ -0,0 +1,31 @@
+#!/usr/bin/env python3
+"""Debug script to trace resolved truth data."""
+
+from gibil.classes.env_loader import EnvLoader
+from gibil.classes.weather_store import WeatherStore
+from gibil.classes.weather_display import WeatherDisplay
+from datetime import datetime, timezone
+
+EnvLoader().load()
+
+store = WeatherStore.from_env()
+dataset = store.load_display_dataset()
+
+print(f"\n=== DEBUG OUTPUT ===")
+print(f"Forecast points: {len(dataset.forecast_points)}")
+print(f"Resolved truth points: {len(dataset.resolved_truth)}")
+
+print(f"\nResolved truth details:")
+for i, point in enumerate(dataset.resolved_truth):
+    print(f"  [{i}] resolved_at={point.resolved_at}, temp={point.temperature_c}, radiation={point.shortwave_radiation_w_m2}")
+
+print(f"\nAPI payload:")
+display = WeatherDisplay()
+payload = display.data_payload(dataset)
+import json
+data = json.loads(payload)
+print(f"  Resolved truth in payload: {len(data['resolved_truth'])}")
+for i, point in enumerate(data['resolved_truth']):
+    print(f"  [{i}] resolved_at={point['resolved_at']}, temp={point['temperature_c']}")
+
+print(f"\n=== END DEBUG ===\n")