Fix runtime restart loops and HF cache conflicts v0.7.4

Fix final compilation issues v0.7.3

Fix remaining moshi API issues v0.7.2

Fix moshi API compilation errors v0.7.1

Implement Plan A+: Real moshi streaming STT pipeline v0.7.0

Fix Send trait error by temporarily removing StreamingModule

Fix Send trait error for StreamingModule in async context

Fix compilation errors while preserving StreamingModule for streaming

Fix model loading error: use STT streaming instead of Mimi codec

Fix moshi API compilation errors for StreamTensor and model loading

Implement real speech transcription with Mimi audio tokenizer

Fix State::default() compilation error and clean up unused imports

Fix compilation errors with working moshi ASR API implementation

Implement real moshi ASR model integration for speech-to-text transcription

fix: Add retry logic for HuggingFace model cache lock acquisition failures

fix: Resolve Rust borrow checker error in model transcribe function

fix: Replace moshi Model/Config imports with placeholder structures

fix: Resolve all Rust compilation errors

Initial deployment of Kyutai STT Server v3