AtropinolTT/dsh-guide-dog: Adding Multimodal Generation and Visual Inspection to the DSH Web Frontend

`AtropinolTT/dsh-guide-dog` is a multimodal plugin for the DSH web profile, built on MiniMax capabilities, designed to integrate visual inspection, media generation, voice interaction, and Web UI preview functionality. Core features include: utilizing VLM for image description and structured inspection (e.g., screenshots, charts); supporting the generation of images, videos, speech, music, and text; providing a unified media path for real-time preview in conversation cards; and supporting real-time call mode with VAD, streaming TTS, and barge-in functionality. The plugin automatically invokes visual tools via system prompts and allows reuse of existing voice skill pipelines, with automatic fallback when missing. Installation requires the `dsh plugin --profile web add` command or static b