Centralized runtime config defaults

All config values are centralized for a more organized functionality.

Wake cues for different aspects of voice input were added.

Tools were added for Text To Speech benchmarking.

All used models during a session are now warmed up from start for faster responses.