Dynamic window metrics for cold start scenarios

Calculate metrics from first relevant event to now (capped at 30 days)
instead of a fixed 30-day window. This fixes inaccurate metrics for new
users who have only a few days of data.

Changes:
- Add _get_first_event_ts() and _calculate_window() helpers to stats.py
- Add window_days field to EggStats dataclass
- Update routes/eggs.py and routes/feed.py to use dynamic window
- Update templates to display "N-day avg" instead of "30-day avg"
- Use ceiling division for window_days to ensure first event is included

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

tests/test_service_stats.py

@@ -489,7 +489,7 @@ class TestEggStatsCaching:
     def test_cached_stats_have_window_bounds(self, seeded_db, e2e_test1_setup):
         """Cached stats include window_start_utc and window_end_utc."""
         ts_utc = e2e_test1_setup["ts_utc"]
-        get_egg_stats(seeded_db, e2e_test1_setup["location_id"], ts_utc)
+        stats = get_egg_stats(seeded_db, e2e_test1_setup["location_id"], ts_utc)
 
         row = seeded_db.execute(
             """
@@ -500,7 +500,6 @@ class TestEggStatsCaching:
         ).fetchone()
 
         assert row is not None
-        assert row[1] == ts_utc  # window_end_utc
-        # Window is 30 days
-        thirty_days_ms = 30 * 24 * 60 * 60 * 1000
-        assert row[0] == ts_utc - thirty_days_ms  # window_start_utc
+        # Cached bounds should match what get_egg_stats returned
+        assert row[0] == stats.window_start_utc
+        assert row[1] == stats.window_end_utc