Bug: unique_values/unique_counts: NaNs should not be treated as equal

[crusaderky]

sparse version checks

• I checked that this issue has not been reported before list of issues.
• I have confirmed this bug exists on the latest version of sparse.
• I have confirmed this bug exists on the main branch of sparse.

Describe the bug

The Array API specification states for unique_* functions:

As nan values compare as False, nan values should be considered distinct.

In other words, it dictates behaviour equal to np.unique(..., equal_nan=False).

sparse.unique_* functions however treat NaNs as equal.

Steps or code to reproduce the bug

>>> from math import nan

>>> import array_api_strict as xp
>>> xp.unique_values(xp.asarray([math.nan, math.nan]))
Array([nan, nan], dtype=array_api_strict.float64)

>>> import sparse
>>> sparse.unique_values(sparse.asarray([math.nan, math.nan]))
array([nan])

Please describe your system.

sparse 0.16.0b4

[hameerabbasi]

NaN handling in sparse arrays is weird, in general. I'm not sure this is fixable within the realm of sparse arrays.