Add multimodal datasets

Epic issue to track the addition of multimodal datasets:
- [ ] #2526
- [x] #1842
- [ ] #1810

Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).

@VictorSanh feel free to add and sort by priority any interesting dataset. I have added the multimodal dataset requests which were already present as issues.