G5: a still asked for at its own size is the unsized frame (#1812)

Review: frame_at_size at a still's native size made a second cache key, so the pool decoded
and kept a second full-size copy of a frame thumbnail() or frame() already had.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

crates/media/src/pool.rs

@@ -716,7 +716,8 @@ impl Inner {
         }
         let mut loc = Self::locate(footage, t, float);
         let movie = loc.media_t.is_some();
-        if let Some(size) = size {
+        // A still asked for at its own size is the unsized frame (one decode, one entry).
+        if let Some(size) = size.filter(|s| movie || *s != (footage.width, footage.height)) {
             loc.key.size = size;
         }
         let mut c = lock(&self.cache);

crates/media/tests/footage.rs

@@ -595,6 +595,11 @@ fn stills_at_reduced_sizes_are_resampled_once() {
     let f = probe(&p).expect("probe");
     let pool = MediaPool::new();
     let full = pool.frame(ItemId(1), &f, Tick::ZERO).expect("frame");
+    // At its own size it is the unsized frame: no second decode, no second full-size copy.
+    let misses = pool.stats().misses;
+    let native = pool.frame_at_size(ItemId(1), &f, Tick::ZERO, f.width, f.height).expect("frame");
+    assert!(Arc::ptr_eq(&full, &native), "the native size reuses the unsized frame");
+    assert_eq!(pool.stats().misses, misses);
     for (w, h) in [(306, 48), (153, 24), (77, 12)] {
         let sized = pool.frame_at_size(ItemId(1), &f, Tick::ZERO, w, h).expect("frame");
         assert_eq!(*sized, effectcraft_raster::resample(&full, w, h), "{w}×{h}: the renderer's resample");