`dpt.unique_inverse` and `dpt.unique_all` returns `inverse_indices` array of int32 dtype

[antonwolfy]

The issue is quite similar to gh-1322.

The result type of inverse_indices may vary on the same platform:

import dpctl, dpctl.tensor as dpt, numpy

dpctl.__version__
# Out: '0.18.0dev0+80.gb7fd613acc'

a = dpt.ones(10, dtype="i8")

dpt.unique_inverse(a).inverse_indices.dtype
# Out: dtype('int32')

dpt.unique_all(a).inverse_indices.dtype
# Out: dtype('int32')

a = dpt.arange(10, dtype="i8")
dpt.unique_inverse(a).inverse_indices.dtype
# Out: dtype('int64')

dpt.unique_all(a).inverse_indices.dtype
# Out: dtype('int64')

while the specification states:

second element must have the field name inverse_indices and must be an array containing the indices of values that reconstruct x. The array must have the same shape as x and have the default array index data type.

and here is about the default array index type:

a default array index data type (either int32 or int64).
...
The default array index data type may be int32 on 32-bit platforms, but the default should be int64 otherwise.