Fix dpctl.tensor.sum for support zero-dimensional array

[vlad-perevezentsev]

This PR changes a bit the behavior for dpctl.tensor.sum function for scalars.
The previous implementation always returned dpt.zeros for zero-dimensions of the input array and worked incorrectly for scalars. According to the Python Array API: If N is 0, the sum is 0 (i.e., the empty sum) [sum].
Now if dpctl.tensor.sum takes a scalar it returns a scalar has the same behavior as numpy

#old impl
a = dpt.asarray(1)
dpt.sum(a)
usm_ndarray(0., dtype=float32)

a = dpt.asarray(())
dpt.sum(a)
usm_ndarray(0., dtype=float32)

#new impl
a = dpt.asarray(1)
dpt.sum(a)
usm_ndarray(1., dtype=float32)

a = dpt.asarray(())
dpt.sum(a)
usm_ndarray(0., dtype=float32)

• Have you provided a meaningful PR description?
• Have you added a test, reproducer or referred to an issue with a reproducer?
• Have you tested your changes locally for CPU and GPU devices?
• Have you made sure that new changes do not introduce compiler warnings?
• Have you checked performance impact of proposed changes?
• If this PR is a work in progress, are you opening the PR as a draft?

[coveralls]

coverage: 84.1% (+0.02%) from 84.079% when pulling 9579d8a on vlad-perevezentsev:fix_dpctl_sum into 47466ee on IntelPython:master.

[ndgrigorian]

LGTM!

[github-actions]

Array API standard conformance tests for dpctl=0.14.3dev3=py310h7bf5fec_33 ran successfully.
Passed: 348
Failed: 652
Skipped: 119