Improve the documentation of the skrub.datasets fetch functions

[rcap107]

At the moment, each dataset has its own version of the documentation. Some have info about the dataset, some have a description of the Bunch object. In general, it's messy and not very uniform.

I think all datasets should include information on the size of the dataset itself (number of rows/columns, size on disk), a description of the task, and in general the format should be standardized.