Return matching_records list from tracks_with_record_info_query

Replace scalar collected_record_id/wishlisted_record_id subqueries with
a single matching_records JSON array grouped by release group
(records.musicbrainz_id). Add parse_matching_records/1 for JSON parsing
and derive_legacy_record_ids/2 as a temporary bridge for existing
templates. Update tests to use shared musicbrainz_id fixtures.

lib/music_library/listening_stats.ex

@@ -356,8 +356,12 @@ defmodule MusicLibrary.ListeningStats do
             'cover_hash', r.cover_hash\
             )) \
             FROM records r \
-            JOIN record_releases rr ON rr.record_id = r.id \
-            WHERE rr.release_id = (? ->> '$.musicbrainz_id'))\
+            WHERE r.musicbrainz_id = (\
+            SELECT r2.musicbrainz_id FROM records r2 \
+            INNER JOIN record_releases rr ON rr.record_id = r2.id \
+            WHERE rr.release_id = (? ->> '$.musicbrainz_id') \
+            LIMIT 1\
+            ))\
             """,
             t.album
           ),
@@ -365,8 +369,10 @@ defmodule MusicLibrary.ListeningStats do
           fragment(
             """
             (SELECT min(ar.musicbrainz_id) FROM artist_records ar \
-            WHERE ar.record_id = (\
-            SELECT rr.record_id FROM record_releases rr \
+            INNER JOIN records r ON r.id = ar.record_id \
+            WHERE r.musicbrainz_id = (\
+            SELECT r2.musicbrainz_id FROM records r2 \
+            INNER JOIN record_releases rr ON rr.record_id = r2.id \
             WHERE rr.release_id = (? ->> '$.musicbrainz_id') \
             LIMIT 1\
             ))\
@@ -377,10 +383,13 @@ defmodule MusicLibrary.ListeningStats do
           fragment(
             """
             (SELECT r.cover_hash FROM records r \
-            JOIN record_releases rr ON rr.record_id = r.id \
+            WHERE r.musicbrainz_id = (\
+            SELECT r2.musicbrainz_id FROM records r2 \
+            INNER JOIN record_releases rr ON rr.record_id = r2.id \
             WHERE rr.release_id = (? ->> '$.musicbrainz_id') \
-            AND r.purchased_at IS NOT NULL \
-            ORDER BY r.id \
+            LIMIT 1\
+            ) \
+            ORDER BY (CASE WHEN r.purchased_at IS NOT NULL THEN 0 ELSE 1 END), r.id \
             LIMIT 1)\
             """,
             t.album

test/music_library/listening_stats_test.exs

@@ -335,16 +335,28 @@ defmodule MusicLibrary.ListeningStatsTest do
     end
   end
 
-  describe "deduplication when multiple records share release IDs" do
+  describe "deduplication when multiple records share a release group" do
     setup do
-      # Create two collected records that share the same MusicBrainz release IDs
-      # (both use the default marbles fixture with identical musicbrainz_data).
-      # This produces two rows per release_id in record_releases, which previously
-      # caused fan-out in ListeningStats joins.
-      record_a = RecordsFixtures.record(%{title: "Marbles Copy A"})
-      record_b = RecordsFixtures.record(%{title: "Marbles Copy B"})
+      # Create two collected records that share the same release group
+      # (musicbrainz_id) but may have different physical release IDs.
+      # The matching_records query groups by release group, so both records
+      # should appear for any scrobble matching either record's release IDs.
+      shared_musicbrainz_id = Ecto.UUID.generate()
 
-      # Pick a release_id shared by both records
+      record_a =
+        RecordsFixtures.record(%{
+          title: "Marbles Copy A",
+          musicbrainz_id: shared_musicbrainz_id
+        })
+
+      record_b =
+        RecordsFixtures.record(%{
+          title: "Marbles Copy B",
+          musicbrainz_id: shared_musicbrainz_id
+        })
+
+      # Pick a release_id from the marbles fixture (shared by both records
+      # since they use the same musicbrainz_data)
       shared_release_id = "d3f9b9e2-73f5-4b47-a2a7-2c2199aad608"
       now = System.system_time(:second)
 
@@ -364,10 +376,11 @@ defmodule MusicLibrary.ListeningStatsTest do
         scrobbled_at_uts: now - 200
       })
 
-      # The lexicographically smaller record_id should consistently win (MIN)
       expected_record_id = Enum.min([record_a.id, record_b.id])
 
       %{
+        record_a: record_a,
+        record_b: record_b,
         expected_record_id: expected_record_id,
         shared_release_id: shared_release_id
       }