dsh-browser-vision: DeepSeek Harness Browser Vision Plugin

`dsh-browser-vision` is a browser vision plugin for DeepSeek Harness, designed to address visual elements that cannot be captured by DOM parsing. It drives a real Chrome browser through CDP and uses a vision model to read Canvas text, charts, and rendered content. The plugin supports page driving, schema-based structured JSON extraction, Markdown output, and cost tracking. Installation requires configuring the Python engine, environment variables, and Chrome, and it provides tools such as `browser_vision_task` and `browser_vision_extract`. When using it, be mindful of the Python version requirements and screenshot parameter settings to help agents process complex visual information more effectively.