For some applications, users may want to explicitly indicate start and end of the user's turn, instead of having it to be a natural conversation.

In this case, the flow might look like this:

• to signal start of user's turn: unmute microphone
• to signal end of user's turn: mute microphone

on the agent side, we would want an option like:

• end_of_turn="manual"
• agent.start_user_turn() and agent_end_user_turn()

while the user's turn is active, the mic's input will be fed into the agent's input. We would want this capability to work the same way between VoicePipelineAgent and MultimodelAgent