feat: add DingTalk media message support (image, file, voice, video)

[nap-liu]

Summary

• Receive: download images/files/audio/video via DingTalk downloadCode API, save to workspace
• Vision: encode images as base64, inject into LLM context with [image_data:] markers
• Rich text: parse richText messages with inline images
• Send back: _upload_dingtalk_media() + _send_dingtalk_media_message() for agent-initiated media
• Thinking reaction: visual indicator while processing (add on receive, recall after reply)
• History: rehydrate_image_messages() re-injects images from disk for multi-turn context
• Channel file sender: ContextVar integration for agent tools to send files via DingTalk

Test plan

• DingTalk E2E: image received → downloaded → saved → rehydrated in history
• Thinking reaction: added on message, recalled after reply (confirmed in logs)
• Rich text: image extracted and processed