dsh-pseudo-vision: Supply local visual evidence to DeepSeek Harness's text-only provider

dsh-pseudo-vision is a plugin designed to provide local visual capabilities to DeepSeek Harness (DSH) text-only models. By converting image attachments into OCR text, color statistics, pixel scanning, and metadata, it enables providers that do not support image input to process visual content based on this "evidence," operating entirely locally without any external API dependencies. The plugin supports generating sibling routes for specific or all providers and features a round-robin hierarchical evidence management system alongside request quota protection mechanisms. Installation requires Node.js >= 22.19 and pnpm, supports deployment via GitHub or local paths, and offers flexible configuration, making it suitable for developers seeking to enhance DSH's multimodal processing capabilities without introducing external visual services.