[feature] WebAI - Add support for SmolVLM 256M (& 500M)

[addyosmani]

SmolVLM 256M (& 500M): The world’s smallest multimodal model. Designed for efficiency and perfect for on-device applications

Demo: https://huggingface.co/spaces/HuggingFaceTB/SmolVLM-256M-Instruct-WebGPU
Twitter: https://x.com/xenovacom/status/1882435994160447587
Examples: https://github.com/huggingface/transformers.js-examples/tree/main/smolvlm-webgpu

It would be great to extend our models with SmolVLM.

There are some UX improvements we could also make at the same time such as, when you use a prompt with an image:

Fixing the [Object] text following a selected image.

Double-checking there are no specific issues with multi-modal support. I was running into issues with our other multimodal model (couldn't get it working), unsure if this was model specific.

[jakobhoeg]

Added a fix to the [Object] text following a selected image.

Also, the multi-modal issue is related to the web-llm package, as mentioned here:
mlc-ai/web-llm#657
mlc-ai/web-llm#640