composite: switch to smooth synthetic atmospheric glow, eliminating foil artefact

The previous approach tried to preserve real sky-cam pixels near the moon and
darken them, but with a bright/hazy sky (thick clouds, full moon glow) the
cloud texture still bled through as the "wrinkled silver foil" look.

New approach in _atmospheric_sky_background():
- Sample the mean atmospheric colour from a ring at 1.15–1.8× moon radius.
- Scale it to a tasteful glow level (×0.35) — real colour, not raw brightness.
- Build a smooth quadratic radial gradient: halo colour at the disk edge,
  deep night-sky (R4 G6 B14) at 3× radius and beyond.

No raw sky-cam pixels appear in the background, so the result is texture-free
regardless of cloud cover or haze. Verified clean against both a normal frame
and a 2.5× brightened worst-case bright-sky simulation.

https://claude.ai/code/session_01HuJ83KvMvshiY6HxJbtMsc

moon_composite.py

@@ -110,36 +110,40 @@ def _atmospheric_sky_background(
     moon_cy: float,
     moon_diam_px: float,
 ) -> Image.Image:
-    """Real atmospheric halo on a deep night-sky backdrop.
+    """Smooth atmospheric glow on a deep night-sky backdrop.
 
-    Inside ~1.8× the moon radius the real sky is preserved (gamma-darkened to
-    collapse diffuse cloud texture while the bright atmospheric glow survives).
-    Beyond ~4.5× the radius it fades smoothly to a near-black night colour, so
-    the result looks like a genuine dark-sky photograph rather than an
-    over-exposed, heavily upscaled camera frame.
+    Samples the real halo colour from a ring just outside the moon disk, then
+    builds a smooth radial gradient from that colour at the disk edge to deep
+    night-sky beyond ~2.5× the radius.  No sky-cam pixels are used directly,
+    so there is no cloud texture or foil artefact regardless of how bright or
+    hazy the original sky was.
     """
     arr = np.asarray(sky_img).astype(np.float32)
     h, w = arr.shape[:2]
-
-    # Deep night-sky colour — very dark desaturated blue
-    night = np.array([4, 6, 14], dtype=np.float32)
-    dark_bg = np.broadcast_to(night, (h, w, 3)).copy()
+    r = moon_diam_px / 2.0
 
     yy, xx = np.ogrid[0:h, 0:w]
     dist = np.sqrt((xx - moon_cx) ** 2 + (yy - moon_cy) ** 2)
 
-    r = moon_diam_px / 2.0
-    inner_r = 1.3 * r   # real-sky halo fully preserved inside here
-    outer_r = 2.2 * r   # fully dark outside here — keeps frame corners black
+    # Sample mean colour of the real halo in a ring from 1.15× to 1.8× radius.
+    ring = (dist >= 1.15 * r) & (dist < 1.8 * r)
+    if ring.any():
+        halo_rgb = arr[ring].mean(axis=0)          # real atmospheric colour
+    else:
+        halo_rgb = np.array([80, 85, 95], dtype=np.float32)
+    # Scale to a tasteful glow level (original is often overexposed)
+    halo_rgb = np.clip(halo_rgb * 0.35, 0, 180).astype(np.float32)
 
-    alpha = np.clip((outer_r - dist) / (outer_r - inner_r), 0.0, 1.0)[..., None]
+    # Deep night-sky colour — very dark desaturated blue
+    night = np.array([4, 6, 14], dtype=np.float32)
 
-    # Gamma + scale: compresses diffuse cloud texture toward black while the
-    # bright atmospheric halo (already near-white) survives nearly intact.
-    real_darkened = np.power(np.clip(arr / 255.0, 0, 1), 1.6) * 0.65 * 255.0
+    # Radial alpha: 1.0 at the moon disk edge (dist_norm=1), 0 at 3× radius.
+    dist_norm = dist / r
+    glow_alpha = np.clip(1.0 - (dist_norm - 1.0) / 2.0, 0.0, 1.0) ** 2
+    glow_alpha = glow_alpha[..., None]
 
-    blended = alpha * real_darkened + (1.0 - alpha) * dark_bg
-    return Image.fromarray(np.clip(blended, 0, 255).astype(np.uint8))
+    bg_arr = glow_alpha * halo_rgb + (1.0 - glow_alpha) * night
+    return Image.fromarray(np.clip(bg_arr, 0, 255).astype(np.uint8))
 
 
 def _disk_mask(size: int, feather_px: int = 6) -> Image.Image: