v0.5.0

## Highlights

- Add Apple Silicon support for the experimental GGML backend using the Metal wheel of `qwentts-cpp-python`.
- Require `qwentts-cpp-python>=0.4.1` for the GGML extra; 0.4.0 did not support macOS.
- Document the published 0.4.1 CPU, CUDA, and Metal wheel variants and their installation paths.

Install on Apple Silicon with `pip install "faster-qwen3-tts[ggml]"`. The Torch backend remains CUDA based; choose `--backend ggml` to use Metal.

## What's changed

- [Support Apple Silicon Metal GGML wheels](https://github.com/andimarafioti/faster-qwen3-tts/pull/138)
- [Prepare release 0.5.0](https://github.com/andimarafioti/faster-qwen3-tts/pull/139)

[Full changelog](https://github.com/andimarafioti/faster-qwen3-tts/compare/v0.4.0...v0.5.0)