Add Audio Response for LLM generated content

[vivekuppal]

Feature Enhancements

• Add Speech Response mode, allowing Transcribe to respond with audio for this conversation.

Bug Fixes / Improvements

• Improve Mic transcription when using a non default mic
• Make Custom Prompts configurable by non developers. Added them to parameters.yaml file
• Remove redundant code related to prompt preambles
• Optimization for the case when LLM is disabled. Skip unnecessary processing.
• Fix incorrect syntax of default system message
• Default system message is also tracked in conversation object with rest of the conversation.

Experimental

• Improvements to multiturn prompts, responses. Response in this case are still erratic at times and need more work.