v2.3.6

• Bump dependencies.
• Fix broken latin-1 parsing:

Latin-1 code points were parsed as utf-8 resulting in the replacement character '�' (U+FFFD).

The program now validates utf-8 tags when parsing. If validation fails the tag is parsed as latin-1.