DeepSeek Harness Plugin: Make DSH Voice

`make-dsh-voice` is a plugin designed for the DeepSeek Harness web interface to address the absence of voice interaction. Based on Alibaba Cloud's CosyVoice model, it provides a `voice_speak` tool, a persistent player, and a voice settings page, with support for configuring API keys, reply strategies, and startup sounds. It is installed via CLI commands, requires Python 3 and the DashScope SDK, and has been tested on Windows 11. It also includes a standalone synthesis script for local testing.