dsh-llm-deepseek-vision: Provide visual input to pure text models in DeepSeek Harness

dsh-llm-deepseek-vision is an MIT-licensed DeepSeek Harness (DSH) plugin designed to address the issue where pure text models cannot directly process images. This plugin registers a `DeepSeek (Vision)` route that utilizes a vision model to convert images into text descriptions, which are then passed to a text model (such as deepseek-v4-flash/pro) for inference. Its working principle involves request-based routing: when there is no image, it is transparently forwarded to the text route; when an image is present, the configured vision model generates a description, and the delegated route performs the inference. Installation requires adding it to the web profile via npm and configuring the visual provider, model, prompt, and delegated route in cordis.patch.yml. Before enabling, it is necessary to ensure the visual model is...