Add support for extracting and saving generated images from OpenRouter
models that return images (like gpt-5-image-mini).
- Extract base64 images from assistant responses via extractAssistantImages()
- Save generated images using standard media store ("generated" subdir)
- Send generated images to users via message channels
- Support both direct agent responses and subagent image generation
This enables use of OpenRouter's image generation models in Moltbot
conversations and subagent tasks.
