added test_model to cli, fixed retained audio cut short, improved "gold ... cmd" recognition accuracy

[jwebmeister]

Tacspeak

• added --test_model to cli.py
• added --transcribe_wav to cli.py
• updated dragonfly version to fix retained audio being cut short
  • may or may not affect speech recognition accuracy

Ready or Not grammar module

• changed many Choice with dummy values, to an Optional with a default value wrapping a Choice
  • improves recognition accuracy of "gold ... cmd", and probably also other cmds
• added "kick the door down"