Made dictation concurrent with token generation

[Jiminator]

Description

Dictation of the voice chat will no longer wait for the model to finish generating a response but instead will be concurrent to token generation. This is done by using multithreading and separating the output into phrases separated by delimiters such as punctuation.

Changes Made

• updated voicechat.cc to accommodate changes made in previous commits.
• changed the generate function to only store phrases of the output instead of the whole response, and to dictate the phrase as soon as it is done generating.
• Uses ", ", "\n", ". ", "! ", "?", ": " as delimiters
• Used multithreading so that generation would not be interrupted by dictation.

Video

https://www.youtube.com/watch?v=mv2y9HRS1mE&feature=youtu.be

Checklist

Please review and check off these items before submitting your pull request:

• [x] I have tested the changes locally, and they work as expected.

Reviewers

@RaymondWang0