gemini-live-2.5-flash-preview-native-audio slow warm up

### Description of the bug:

I have two issues with the model on Pipecat:
[CA1b7409a900b00e654b7058c97dcbcd7a.mp3](https://github.com/user-attachments/files/22983636/CA1b7409a900b00e654b7058c97dcbcd7a.mp3)

1.The first response is very slow. Is there an chance to warm up the connection for the new native-audio model?

2. The model, seems to shoot multiple answers, without waiting for the answer really. For example:

- "..How are you Today?
- "Great, I will be quick.." is added right after

In theory it should have waited for the user answering first. Somehow this issue happen more in the beginning of the conversation.

### Actual vs expected behavior:

_No response_

### Any other information you'd like to share?

_No response_