1.2.4

• Words are handled as characters instead of bytes in the parser, so multibyte text is no longer split in the middle of a character. This fixes keywords containing a stray question mark, such as «москва? (#21, #7)
• The minimum word length is counted in characters instead of bytes, so short words are now filtered the same way in every language
• Added tests for keyword encoding, the character based minimum length, and sentence punctuation
• Latin text is unaffected, the output is unchanged for English.