Fix memory leak in CachedGenreRepository (#3031)

that the scanner was run, the ttlcache was also created each time.
This caused (under testing with 166 genres in the database) the
memory consumed by navidrome to 101.18MB over approx 3 days; 96%
of which is in instances of this cache. Swapping to a singleton
has reduced this to down to ~ 2.6MB

Co-authored-by: Rob Emery <git@mintsoft.net>

scanner/cached_genre_repository.go

@@ -8,25 +8,27 @@ import (
 	"github.com/jellydator/ttlcache/v2"
 	"github.com/navidrome/navidrome/log"
 	"github.com/navidrome/navidrome/model"
+	"github.com/navidrome/navidrome/utils/singleton"
 )
 
 func newCachedGenreRepository(ctx context.Context, repo model.GenreRepository) model.GenreRepository {
+	return singleton.GetInstance(func() *cachedGenreRepo {
 		r := &cachedGenreRepo{
 			GenreRepository: repo,
 			ctx:             ctx,
 		}
 		genres, err := repo.GetAll()
+
 		if err != nil {
 			log.Error(ctx, "Could not load genres from DB", err)
-		return repo
+			panic(err)
 		}
-
 		r.cache = ttlcache.NewCache()
 		for _, g := range genres {
 			_ = r.cache.Set(strings.ToLower(g.Name), g.ID)
 		}
-
 		return r
+	})
 }
 
 type cachedGenreRepo struct {