minirst: use unicode string as intermediate form for replacement Some character encodings use ASCII characters other than control/alphabet/digit as a part of multi-bytes characters, so direct replacing with such characters on strings in local encoding causes invalid byte sequences.