All dtype field names are already Unicode strings, and we no longer need
to check for byte strings and convert them to Unicode.