Local text-to-speech backend for Blazen using Piper.

This crate provides fully local, offline TTS synthesis via Piper voice models running on ONNX Runtime, with no API keys required.

When used through blazen-llm with the piper feature flag, this crate's [PiperProvider] will implement blazen_llm::AudioGeneration.

Feature flags

Without the engine feature the crate compiles (options struct + stub provider) but cannot actually run synthesis. This keeps workspace builds fast when the heavy native dependencies are not needed.