[Feature Request] Support Local/Offline ASR Voice using Sherpa-ONNX (Rust)

Description: Is your feature request related to a problem? Please describe. Currently, Claurst's voice mode relies on cloud-based ASR (like OpenAI Whisper API), which introduces latency, costs API credits, and requires a stable internet connection. Describe the solution you'd like I would like to propose integrating a local, high-performance ASR engine. Specifically, Sherpa-ONNX seems like a perfect fit because it provides a native Rust API and supports various efficient models (like SenseVoice or Paraformer) that run fast on CPUs. Reference: Sherpa-ONNX Rust API Examples, Fun-asr Describe alternatives you've considered I also looked into FunASR, which has excellent accuracy (especially for multilingual/Chinese support), but it is primarily Python-based. Integrating it into Claurst might introduce heavy runtime dependencies compared to the C++/Rust nature of Sherpa-ONNX. Additional context Integrating Sherpa-ONNX would: Enable fully offline voice-to-text. Drastically reduce latency for voice commands. Improve privacy as no audio data leaves the local machine. I believe this would make Claurst's "Soul" feel much more responsive and "alive."