3.1.0

• implemented a pre-computed decoder plan for each class (#206)
  • improved decoding efficiency by ~20%
• added json_to_data to aid in re-encoding JSON messages
  • decodes the base64 string fields produced by to_json back into bytes
  • thanks to @fabrepe
• encode_msg & encode_dict now require a mandatory sentence argument
  • !BREAKING CHANGE!
  • previously talker_id was used to pass both talker and sentence type
  • this was incorrect and did not match the implementation in messages.py
  • before: `encode_msg(talker_id="AIVDO", ...)