Add __add__(), __iadd__() and similar to Dataset class

[sgraaf]

Is your feature request related to a problem? Please describe.
No.

Describe the solution you'd like
I would like to be able to concatenate datasets as follows:

>>> dataset["train"] += dataset["validation"]

... instead of using concatenate_datasets():

>>> raw_datasets["train"] = concatenate_datasets([raw_datasets["train"], raw_datasets["validation"]])
>>> del raw_datasets["validation"]

Describe alternatives you've considered
Well, I have considered concatenate_datasets() 😀

Additional context
N.a.

[pri1311]

I was going through the codebase, and I believe the implementation of add() and iadd() will be similar to concatenate_datasets() after the elimination of code for arguments other than the list of datasets (info, split, axis).
(Assuming elimination of axis means concatenating over axis 1.)

[mariosasko]

Most data frame libraries (Polars, Pandas, etc.) override __add__ to perform (mathematical) summation, so having different behavior could lead to confusion.