fix: always write artist row in index to prevent redundant re-indexing

indexOneArtist only wrote the artist entry when aliases were non-empty.
Artists without MB aliases (common for smaller/niche artists) never got
an entity_type='artist' row, so indexedArtistMBIDs() couldn't see them.
filterUnindexed then treated them as new on every startup, triggering
redundant LB API calls for top-release-groups and top-recordings.

Now the artist row is always written, with aliases as an empty string
when none exist. Subsequent builds will correctly skip these artists.

backend/explore/searchindex.go

@@ -1195,11 +1195,12 @@ func (si *SearchIndex) indexOneArtist(
 
 	wg.Wait()
 
-	// Extract aliases from the now-cached MB rels (populated by
+	// Write the artist entry into the index so indexedArtistMBIDs()
-	// the image resolution above) and update the artist's index entry.
+	// recognises this artist as processed on subsequent builds.
+	// Also stores aliases from the now-cached MB rels (populated
+	// by the image resolution above) for FTS search.
 	if si.artistImg != nil {
 		aliases := si.artistImg.GetAliases(artist.ArtistMBID)
-		if aliases != "" {
 		si.writeBatch([]SearchIndexResult{{
 			EntityType: "artist",
 			MBID:       artist.ArtistMBID,
@@ -1210,7 +1211,6 @@ func (si *SearchIndex) indexOneArtist(
 			Aliases:    aliases,
 		}})
 	}
-	}
 
 	// Batch write discography results.
 	all := make([]SearchIndexResult, 0, len(rgs)+len(recs))