API client for `datasets-server`

[Wauplin]

Originally from @davanstrien on slack (internal link)

Datasets Server is a lightweight web API for visualizing and exploring all types of datasets stored on the Hugging Face Hub (see repo and docs). Having a client API in huggingface_hub would make sense to lower the barrier.

Example:

from huggingface_hub import datasets_server

rows = datasets_server.get_first_rows('imdb', config='default', n_rows=10) # returns a generator of rows 

columns = datasets_server.get_datasets_features('imdb')

The client could live either in huggingface_hub or datasets. From @davanstrien: My 2c is that it could make sense to have it in huggingface_hub more since this library is already interacting with many APIs related to the hub. I think there could also in future be other endpoints on datasets-server which are dataset related but more on the hub side than related to the dataset itself i.e. find similar datasets based on content (huggingface/dataset-viewer#396)

cc @severo @lhoestq

[davanstrien]

Happy to help with the implementation for this if useful :)

[severo]

The client could live either in huggingface_hub or datasets.

It would fit better here, since Datasets Server only processes Hub's datasets, while the datasets library can handle any kind of data (local, hosted on the Hub, hosted elsewhere)