Only allow DataFrame as data for DataTable

[iron3oxide]

Currently there are multiple ways to give a DataTable its data. It can be a pandas DataFrame, a list of dictionaries and so forth. This means there has to be a not-so-trivial process of figuring out which type data has and then proceeding accordingly, while also checking for conflicts like DataFrame + a columns argument.

While it is great to offer users the flexibility to define their table data how they see fit, I argue that it would be cleaner and less complex to coerce them to use a "real" tabular data structure, the DataFrame. I also implore you to check out polars, a faster pandas alternative which may be an even better fit. DataFrames can easily be created from all kinds of data structures and file types, so it shouldn't be a DX burden.

This way, some code (utils.is_dataframe(), the columns prop and all associated checks) could be deleted and users would have one obvious way to do things. What do you think?

[SLLau]

Yes please to supporting POLARS