visual-review: Enable DSH Web to display images and invoke visual interpretation

visual-review is an MIT-licensed plugin for DSH designed to address the issues where images cannot be displayed directly in Web chat interfaces and text models cannot understand image content. The plugin supports rendering images in formats such as PNG and JPEG within DSH Web, and provides the `visual_review` tool, which returns a Chinese description of the image by invoking a visual multimodal model, covering text, objects, scenes, and other content. The plugin supports a dual-engine mode: it prioritizes OpenAI-compatible cloud-based multimodal APIs, automatically falling back to the local Qwen3-VL-8B model when not configured to ensure data privacy. Users can flexibly view or update cloud configurations (URL, API Key, Model) via the `visual_review_config` tool, or clear the configuration to...