[BUG] DataFrame.describe prints datetimes differently than pandas

[bdice]

Describe the bug
The DataFrame.describe string formatting for datetime differs between cudf and pandas. Specifically, the mean is printed differently. I think the types are the same, and that this is only an issue with formatting -- not data types or casting -- but I'm not completely sure.

Steps/Code to reproduce bug

>>> cudf.Series([
...     np.datetime64("2000-01-01"),
...     np.datetime64("2010-01-01"),
...     np.datetime64("2010-01-01"),
... ]).describe(datetime_is_numeric=True)
count                                3
mean     2006-09-01T08:00:00.000000000
min                2000-01-01 00:00:00
25%                2004-12-31 12:00:00
50%                2010-01-01 00:00:00
75%                2010-01-01 00:00:00
max                2010-01-01 00:00:00
dtype: object

Expected behavior

>>> pd.Series([
...     np.datetime64("2000-01-01"),
...     np.datetime64("2010-01-01"),
...     np.datetime64("2010-01-01"),
... ]).describe(datetime_is_numeric=True)
count                      3
mean     2006-09-01 08:00:00
min      2000-01-01 00:00:00
25%      2004-12-31 12:00:00
50%      2010-01-01 00:00:00
75%      2010-01-01 00:00:00
max      2010-01-01 00:00:00
dtype: object

This came up in doctests (#9815) because the existing formatting of the docstring does not match either pandas or the current cudf implementation.

cudf/python/cudf/cudf/utils/docutils.py

Lines 207 to 225 in 6b579c7

	>>> s = cudf.Series([
	... np.datetime64("2000-01-01"),
	... np.datetime64("2010-01-01"),
	... np.datetime64("2010-01-01")
	... ])
	>>> s
	0 2000-01-01
	1 2010-01-01
	2 2010-01-01
	dtype: datetime64[s]
	>>> s.describe()
	count 3
	mean 2006-09-01 08:00:00.000000000
	min 2000-01-01 00:00:00.000000000
	25% 2004-12-31 12:00:00.000000000
	50% 2010-01-01 00:00:00.000000000
	75% 2010-01-01 00:00:00.000000000
	max 2010-01-01 00:00:00.000000000
	dtype: object

In the short term, I am updating the doctest in #9815 to match the current implementation (5111a71).

[bdice]

This appears to be resolved by #9867. The changes in _describe_timestamp fixed it by calling

"mean": str(pd.Timestamp(self.mean())),

https://github.com/rapidsai/cudf/pull/9867/files#diff-20a3f41cdf918eefd141f3aba57a2db388915724daea1f3829366cbb591e72ceR3403