Automatically load dataset as pandas

[mfeurer]

This issue is a proposal that we (1) load datasets as pandas by default and (2) rewrite the dataset loader to be pandas by default and convert to numpy if the user requests a numpy array.

The reasons for this proposal are:

1. pandas is much more stable as it used to be a few years ago when we started this project and can now also properly handle strings (see Proposal: Use pandas str type for str datasets #1107).
2. pandas can properly encode categorical columns, which can make it easier for projects building on OpenML-Python to handle these categories.
3. We will use parquet in the background to store files anyway, which has to be interfaced with pandas.

[SubhadityaMukherjee]

Should be fixed in #1372

[PGijsbers]

For future reference, we normally close the issue after it is merged :)
Though it is good that you made the link here since that should've been done when opening the PR.
E.g., if you specify "fixes #xyz" as part of the PR text, then it will automatically close the issue if the PR is merged :)

[SubhadityaMukherjee]

Ah I did not know that. Will keep that in mind thanks!