- Writing is slightly faster
with deduplication, it is around ~25% faster, may be more depending on input
- Added 'allowdeduplication' parameter to encoder.read()
- Fixed api documentation in README