Vision docs 📝

[merveenoyan]

This is a big PR to refresh outdated parts of vision docs and add new ones for newer tasks.

Here's some action items:

• add SAM2 FT to mask-generation tutorial
• add trackio to vision examples (requires extended support from Trainer)
• add universal segmentation docs (@ariG23498 is taking it up)
• fix dataset in semantic segmentation doc (there was a canonical one which didn't load, fixed it)
• replace video classification with VJEPA, change video backend (torchvideo tools no longer work)
• (maybe) swap models with newer ones

[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.

[molbap]

Subscribed, thanks for starting this! Feel free to ping me for a review