Support utils.testing.assert_equal to compare NaN value

[corona10]

Description

According to https://bugs.python.org/issue28579 does not work as I expected.
But ML package has a lot of situation to handling NaN value and this situation actually happened
during I worked on #13231 (comment).

scikit-learn/sklearn/utils/testing.py

Line 89 in 7639886

assert_equal = _dummy.assertEqual

So my proposal is implementing new assert_equal to support NaN equal operation.

And if this proposal is accepted, I'd like to work on this issue.

Steps/Code to Reproduce

from sklearn.utils.testing import assert_equal

assert_equal(float('NaN'), float('NaN'))

Expected Results

from sklearn.utils.testing import assert_equal

assert_equal(float('NaN'), float('NaN')) # pass

Actual Results

from sklearn.utils.testing import assert_equal

assert_equal(float('NaN'), float('NaN')) # fail

Versions

Scikit-Learn 0.21.dev0

[jnothman]

We are no longer really using assert_equal, after moving to pytest. The helpers we use to assert equality between arrays do check for equality of NaN positions.

[corona10]

@jnothman Thanks looks like we can close this issue. I will close this issue.