[Feature Request]: Support Third-Party Streaming Speech Recognition APIs

Hi, thank you for the great work on this project. I really like it and it also really help me! I believe adding some feature would significantly improve the overall experience.

### Background

I would like to request support for configurable third-party streaming speech recognition (ASR) APIs, such as [Volcengine API](https://www.volcengine.com/docs/6561/1354867?lang=en), similar to how [Spokenly](https://spokenly.app) allows users to configure external ASR services. Because different models have better support for different languages

### Motivation

The current macOS built-in speech recognition has limited accuracy, especially for non-English languages. In my case (Chinese):

- Recognition accuracy is noticeably lower than expected.
- Real-time follow scrolling is unreliable due to recognition errors.
- Overall usability is reduced in multilingual scenarios.

### Observed Issue

During testing, I also noticed:

- Recording sometimes does not automatically stop at the end of speech.
- This behavior seems more frequent when using Chinese.
- It may be related to how speech boundaries are detected, though I am not certain.

### Proposed Improvement

- Allow users to configure third-party streaming ASR APIs.
- Provide a pluggable architecture for custom ASR backends.

This would:

- Improve multilingual accuracy.
- Enhance real-time responsiveness.
- Provide flexibility for advanced users.

---

Thank you for considering this feature request.