perf(persistence): avoid a full media_file scan when resolving playlist paths

FindByPaths built one OR-ed equality term per path. On the real media_file
schema SQLite abandons the path index at just two OR-ed terms and falls back to
SCAN media_file, re-testing every term against every row, so the cost grows with
(rows x terms).

Group the candidates by library and emit one IN list per library instead, which
plans as SEARCH media_file USING INDEX media_file_path_nocase. The NOCASE
collation is kept so ASCII case-insensitive matching still works.

This is the dominant cost of M3U playlist import, which resolves every track on
every scan. Measured with a 1000-track playlist against a migrated DB:

  100k media_file rows:   397 -> 51,414 tracks/sec
  500k media_file rows:  78.5 -> 47,174 tracks/sec

The rate no longer degrades as the table grows, which is the expected shape for
an index lookup. Reported in #6043, where an 8 hour scan of a 2M-song library
spent 7h52m in the playlist phase.

persistence/mediafile_repository.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"fmt"
 	"iter"
+	"maps"
 	"slices"
 	"strconv"
 	"strings"
@@ -355,7 +356,10 @@ func (r *mediaFileRepository) GetCursorWithArtwork(options ...model.QueryOptions
 // Library-qualified paths search within the specified library, while unqualified paths
 // search across all libraries for backward compatibility.
 func (r *mediaFileRepository) FindByPaths(paths []string) (model.MediaFiles, error) {
-	query := Or{}
+	// One IN list per library instead of one OR term per path: SQLite abandons the
+	// path index at just two OR-ed equality terms and scans the whole table.
+	byLibrary := map[int][]string{}
+	var unqualified []string
 
 	for _, path := range paths {
 		parts := strings.SplitN(path, ":", 2)
@@ -366,17 +370,24 @@ func (r *mediaFileRepository) FindByPaths(paths []string) (model.MediaFiles, err
 				// Invalid format, skip
 				continue
 			}
-			relativePath := parts[1]
-			query = append(query, And{
-				Eq{"path collate nocase": relativePath},
-				Eq{"library_id": libraryID},
-			})
+			byLibrary[libraryID] = append(byLibrary[libraryID], parts[1])
 		} else {
 			// Unqualified path: search across all libraries
-			query = append(query, Eq{"path collate nocase": path})
+			unqualified = append(unqualified, path)
 		}
 	}
 
+	query := Or{}
+	for _, libraryID := range slices.Sorted(maps.Keys(byLibrary)) {
+		query = append(query, And{
+			Eq{"path collate nocase": byLibrary[libraryID]},
+			Eq{"library_id": libraryID},
+		})
+	}
+	if len(unqualified) > 0 {
+		query = append(query, Eq{"path collate nocase": unqualified})
+	}
+
 	if len(query) == 0 {
 		return model.MediaFiles{}, nil
 	}

persistence/mediafile_repository_test.go

@@ -1055,6 +1055,28 @@ var _ = Describe("MediaRepository", func() {
 				Expect(results).To(HaveLen(1))
 				Expect(results[0].ID).To(Equal("otherlib-track"))
 			})
+
+			It("resolves paths from multiple libraries in a single call", func() {
+				adminMr := NewMediaFileRepository(request.WithUser(GinkgoT().Context(), adminUser), GetDBXBuilder())
+				results, err := adminMr.FindByPaths([]string{
+					"1:artist/Album/track.mp3",
+					fmt.Sprintf("%d:hidden/test.mp3", otherLib.ID),
+				})
+				Expect(err).ToNot(HaveOccurred())
+				Expect(results).To(HaveLen(2))
+				Expect([]string{results[0].ID, results[1].ID}).To(ConsistOf("findpath-1", "otherlib-track"))
+			})
+
+			It("keeps each path scoped to its own library when several are queried", func() {
+				adminMr := NewMediaFileRepository(request.WithUser(GinkgoT().Context(), adminUser), GetDBXBuilder())
+				// Each path exists, but under the other library's ID, so neither must match.
+				results, err := adminMr.FindByPaths([]string{
+					fmt.Sprintf("%d:artist/Album/track.mp3", otherLib.ID),
+					"1:hidden/test.mp3",
+				})
+				Expect(err).ToNot(HaveOccurred())
+				Expect(results).To(BeEmpty())
+			})
 		})
 	})