remove vision2seq vs image-text-to-text

[Vaibhavs10]

reference internal chat: https://huggingface.slack.com/archives/C070Q9GGGGY/p1756985150983899

Qwen models are tagged as both vision2seq as well as image-text-to-text, removing the former since it's deprecated.

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: auto

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[zucchini-nlp]

Hm, I think this is quite breaking. The class is deprecated, yet some users can be still using it because we raise warning until v5

Isn't the hub snippet issue fixed in hub side?

[pcuenca]

Oh yes, this is breaking.

Additional context: https://huggingface.co/datasets/huggingface/transformers-metadata/discussions/7

Other alternatives:

• Create a workaround around here.
• Leave it unchanged until v5. As far as I know, generate and the snippet would still work even if the model is instantiated with AutoModelForVision2Seq.