update to 1.5.21
    
    - add BuildRequires:  hunspell-fr (for make check)
    - Handle Return and Enter correctly when the cursor is not at
      the end of the preëdit (Resolves: rhbz#1418313)
    - Values of spin buttons should not be translatable
    - Make the categories from emojione translatable
    - Make emoji matching accent insensitive
    - If available use pykakasi to convert Japanese emoji category
      names to hiragana
    - If available use the “pinyin” Python module to add pinyin
      to the Chinese names and keywords
    - Don’t fallback to “zh” from “zh_TW”, “zh_HK”, “zh_MO” and “zh_Hant”
    - Don’t sort the labels when listing similar emoji
    - Don’t change Unicode categories to lowercase when loading,
      use the original case
    - Also treat categories 'Zl' and 'Zp' as invisible and add
      Unicode code point
    - When searching for similar emoji, the original emoji should be
      most similar to itself
    - Fix duplicate listing of labels when looking up similar emoji
    - Make it optionally possible to match emoji in Japanese using romaji
    - itb_emoji.py: Add the code point to the name of invisible
      characters also when looking up similar characters
    - Better matching of the Unicode categories
    - Small performance optimization in EmojiMatcher.similar()
    - Remove any U+2028 LINE SEPARATOR and U+2029 PARAGRAPH SEPARATOR characters
      from the lookup table
    - Nicer display of the matching labels when looking up similar emoji
    - Don’t strip mathematical symbols (category 'Sm') from tokens
    - Update of en.xml from CLDR’s emoji annotations
    - Update translations from zanata (de, pl updated)