dsha-tts-bridge: Moving DSH's speech synthesis to the host side

dsha-tts-bridge is a host-side voice synthesis bridge for DeepSeek Harness, maintained by flyinghouse-ma. It addresses the issues of inaccessible `/dev/snd` in Android proot containers and the inability of Android WebView's `speechSynthesis` to pronounce Chinese text: it converts text to MP3 using the WebSocket protocol used by Microsoft Edge in the host process, exposing a same-origin route `/dsh-tts/say.mp3` via DSH's built-in webServer, allowing pages to directly reference the audio using `<audio>`. The plugin has zero npm dependencies, featuring a built-in LRU cache (last 40 entries) and a 15-second timeout; it returns `audio/mpeg` on success and 502 on failure. The text parameter has a limit of 200 characters, and the default voice is `zh-CN-XiaoxiaoNeural`. The installation command is `dsh plugin --profile web add @flyinghouse-ma/dsha-tts-bridge`, which requires restarting `dsh web`. The plugin only performs text-to-MP3 conversion; playback control is handled by the page. It requires access to Microsoft's speech endpoint and is licensed under the MIT License.