`dpnp.unique` with `equal_nan=True` and `axis` specified gives incorrect output for an input with NaNs

[ndgrigorian]

More or less identical to NumPy issue: numpy/numpy#23286

import dpnp

a = dpnp.array([0., np.nan, 2., np.nan, 2., 1., 0., 1., 2., 0.])
print(dpnp.unique(a, equal_nan=True, axis=0))
# [ 0.  1.  2. nan nan]

Expected result is [ 0. 1. 2. nan]