Use cached default device in tensor

[oleksandr-pavlyk]

Profiling of an example of GPU Max revealed that repeated calls to dpctl.select_default_device() took 200x longer than actual computatations.

Introduced dpctl._sycl_device_factory._cached_default_device() and deployed it in tensor (in "_usmarray.pyx" and in "_device.py").

• 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?

[github-actions]

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

[coveralls]

coverage: 91.072% (-0.03%) from 91.099%
when pulling fb47db4 on use-cached-default-device-in-tensor
into be4a01c on master.

[github-actions]

Array API standard conformance tests for dpctl=0.17.0dev0=py310h15de555_33 ran successfully.
Passed: 904
Failed: 2
Skipped: 94

[oleksandr-pavlyk]

Verifying on a machine with two GPU Max cards:

In [3]: %timeit foo_dpt(1_000_000, None)
218 µs ± 55.5 µs per loop (mean ± std. dev. of 7 runs, 1 loop each)

In [4]: %timeit foo_dpt(1_000_000, None)
129 µs ± 2.35 µs per loop (mean ± std. dev. of 7 runs, 10,000 loops each)

In [5]: q = dpctl.SyclQueue()

In [6]: %timeit foo_dpt(1_000_000, q)
154 µs ± 29.1 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)

In [7]: %timeit foo_dpt(1_000_000, q)
118 µs ± 15 µs per loop (mean ± std. dev. of 7 runs, 10,000 loops each)

Previously, the timing was

In [3]: %timeit foo_dpt(1_000_000, None)
6.87 ms ± 99.3 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

In [4]: %timeit foo_dpt(1_000_000, None)
7.07 ms ± 18.6 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

In [5]: q = dpctl.SyclQueue()

In [6]: %timeit foo_dpt(1_000_000, q)
280 µs ± 81.3 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)

In [7]: %timeit foo_dpt(1_000_000, q)
248 µs ± 3.45 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)

[ndgrigorian]

LGTM, this is a good change, thank you @oleksandr-pavlyk