[breaking] [py] Drop support for datatable.

[trivialfis]

Related microsoft/LightGBM#6662 .

Not to be confused with the R data.table, this is the Python version.

cc @jameslamb

We are working on auto re-coding for categorical features with dataframe inputs, it's unlikely that we will have the same treatment for datatable. It's better to remove it so that we can have consistent behaviors across all dataframe implementations.

[jameslamb]

Agree, I think it's a good time for this, for all the reasons mentioned in that LightGBM issues you linked to.

I did this just to check:

git grep -i -E 'datatable|h2o'

I see a few other uses that should be removed:

xgboost/python-package/pyproject.toml

Line 46 in 4b2001e

datatable = ["datatable"]

xgboost/python-package/xgboost/testing/__init__.py

Lines 155 to 156 in 4b2001e

	def no_dt() -> PytestSkip:
	return no_mod("datatable")

xgboost/doc/python/python_intro.rst

Lines 169 to 171 in 4b2001e

	| datatable.Frame | T | FF | | NPA | FF | |
	+-------------------------+-----------+-------------------+-----------+-----------+--------------------+-------------+
	| datatable.Table | T | FF | | NPA | FF | |

[trivialfis]

Yup, will prune all the related code and references.

[trivialfis]

Will merge if there's no objection.