Dpctl raises `ValueError` when changing to larger dtype through the constructor

[antonwolfy]

Dpctl does not allow to change dtype to larger one thorugh the constructor call for 0-size array:

import dpctl, dpctl.tensor as dpt

dpctl.__version__
# Out: '0.21.0dev0+49.g5be8ef909c'

# create 0-sized array
a = dpt.zeros((0,), dtype='i2')

b = dpt.usm_ndarray((0,), dtype='i1', buffer=a)  # it works
c = dpt.usm_ndarray((0,), dtype='i4', buffer=a)  # it fails
ValueError                                Traceback (most recent call last)
Cell In[9], line 1
----> 1 c = dpt.usm_ndarray((0,), dtype='i4', buffer=a)

File dpctl/tensor/_usmarray.pyx:388, in dpctl.tensor._usmarray.usm_ndarray.__cinit__()

ValueError: buffer='[]' can not accommodate the requested array.

The input array has zero size, so it is expected that it can be casted to any data type.