Add dpctl.SyclQueue.memset() method

[vlad-perevezentsev]

This PR proposes adding dpctl.SyclQueue.memset and dpctl.SyclQueue.memset_async methods as a Python wrapper over sycl::queue::memset backed by the existing DPCTLQueue_Memset and a new DPCTLQueue_MemsetWithEvents C-API function and adding a new test_sycl_queue_memset.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?
• 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]

View rendered docs @ https://intelpython.github.io/dpctl/pulls/2361/index.html

[vlad-perevezentsev]

@ndgrigorian I implemented cpdef memset without validating val, i.e. matching sycl::queue::memset semantics where the value is silently truncated toits least significant byte.
Does it make sense to explicitly validate that val is within [0, 255]?

[coveralls]

Coverage is 74.51% — add_sycl_queue_memset into master. No base build found for master.

[antonwolfy]

The docstring promises val is interpreted as unsigned char (LSB only), but nothing verifies e.g. memset(mem, 0x1FF) fills 0xFF or -1 fills 0xFF. There is no error case to test here (values truncate, not reject).

[vlad-perevezentsev]

This case is already covered by test_memset_value_truncated_to_byte (negatives and overflow)
https://github.com/IntelPython/dpctl/pull/2361/changes#diff-8cf72656d339e0f446c2227869696dee611bdce0bc37434f62419482dc4cd075R192

[antonwolfy]

e2 overwrites the entire buffer with 0x02, the final state is 0x02 no matter what e1 did or whether e2 actually waited for it.
To actually prove the dependency you'd need the two ops to interact where order changes the value — e.g. e1 fills the whole buffer with 0x01, e2 (depending on e1) fills only the first half with 0x02, then assert first half 0x02 + second half 0x01.

[antonwolfy]

Not every SYCL device supports every USM kind. Needs to check the aspect (or catch the allocation failure) and pytest.skip when unsupported.
Applicable to the memcpy tests also.

[antonwolfy]

Might raise OverflowError also:

import dpctl
import dpctl.memory as dpm

q = dpctl.SyclQueue()
m = dpm.MemoryUSMShared(16, queue=q)

# val too big for a C int (typically 32-bit here):
q.memset(m, 2**40)
# OverflowError: Python int too large to convert to C long

# negative count into an unsigned size_t:
q.memset(m, 0xAB, -1)
# OverflowError: can't convert negative value to size_t

[antonwolfy]

Suggested change

	return wrap<event>(new event(ev));
	return wrap<event>(new event(std::move(ev)));