Move to Go 1.27, and fix initials for non-ASCII names

Initials() capped the loop with len(out) == 2, which counts bytes: a name
starting with Ä, Ö or Å filled the budget on its own and returned a single
letter. Count the initials taken instead.

go fix also wanted a strings.Builder here, but that allocates a string per
iteration to measure a two-character result. Took its SplitSeq suggestion
in lyrics.go, which drops an intermediate slice.

lyrics.go

@@ -67,7 +67,7 @@ func parseLRC(s string) []lyricLine {
 		return nil
 	}
 	var out []lyricLine
-	for _, raw := range strings.Split(s, "\n") {
+	for raw := range strings.SplitSeq(s, "\n") {
 		stamps := lrcOne.FindAllStringSubmatch(raw, -1)
 		if len(stamps) == 0 {
 			continue