Fix incorrect tests in test_usm_ndarray_dlpack

[ndgrigorian]

The test test_dlpack_device attempted to query a device using DLPack device ID from the list of root devices created by all_root_devices fixture in test_usm_ndarray_dlpack.

This logic is incorrect, as all_root_devices caches only the first two root devices per platform. As the DLPack device ID is based on the ordinal ID of the device as it is canonically presented in get_devices, this could fail if a given platform had more than 2 associated devices (in this case, for a machine with 2 PVC. Each has 2 stacks, and therefore, 4 level_zero and opencl:gpu devices).

• 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?
• Have you added documentation for your changes, if necessary?
• Have you added your changes to the changelog?
• If this PR is a work in progress, are you opening the PR as a draft?

[github-actions]

Deleted rendered PR docs from intelpython.github.com/dpctl, latest should be updated shortly. 🤞

[github-actions]

Array API standard conformance tests for dpctl=0.19.0dev0=py310h93fe807_503 ran successfully.
Passed: 894
Failed: 2
Skipped: 118

[coveralls]

coverage: 88.24%. remained the same
when pulling 7ef0e44 on fix-dlpack-test
into bfd4d57 on master.

[oleksandr-pavlyk]

Yes, you're right. Expect this change to slow down run-time of this test file on the machine with many cards, but correctness first.

Was it failing in Tiber?

[ndgrigorian]

Yes, you're right. Expect this change to slow down run-time of this test file on the machine with many cards, but correctness first.

Was it failing in Tiber?

It was during a test run on a machine with two 1550 Max cards.