Set default dtype of usm_ndarray depending on capabilities of device

[vlad-perevezentsev]

This PR solves the issue of creating usm_ndarray without dtype argument. The default value of dtype in __cinit__ of usm_ndarray is defined as dtype='f8' which raised an exception on Iris Xe which does not support 64-bit precision floating point type.

import dpctl.tensor as dpt

a = dpt.usm_ndarray(())
ValueError: Device Intel(R) Graphics [0x46a6] does not support f8 natively.

Idea of this PR is to change default value dtype="f8" to dtype=None and return default floating point type supported by the device when dtype=None.

import dpctl.tensor as dpt

dpt.usm_ndarray(())
usm_ndarray(0., dtype=float32)

dpt.usm_ndarray((), dtype="f8")
ValueError: Device Intel(R) Graphics [0x46a6] does not support f8 natively.

Note:
I'm not sure if it's correct to get device using dpctl.SyclDevice() when queue is unknown.

• 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: 83.273% (-0.01%) from 83.287% when pulling 2c137e4 on vlad-perevezentsev:flexable_usmarray_type into bc59de9 on IntelPython:master.

[ndgrigorian]

I think it may be more correct to use dpctl.select_default_device here rather than dpctl.SyclDevice, it makes the intent more clear.

[ndgrigorian]

LGTM
Thanks @vlad-perevezentsev!

[github-actions]

Array API standard conformance tests for dpctl=0.14.5dev0=py310h7bf5fec_14 ran successfully.
Passed: 448
Failed: 552
Skipped: 119

[oleksandr-pavlyk]

Thank you @vlad-perevezentsev, great addition!