1.2.1
  - Fix an UBSAN problem in the example
  - Do not run tests if 'readtext' is not unavailable (BDR)

1.1
  - detect_language() is now vectorized
  - detect_language() now returns codes by default to be compatible with 'cld3'
  - detect_language_multi() has been renamed to detect_language_mixed()

1.0
  - Initial CRAN release
