test(cuda.core): request init_cuda for test_latchkernel

test_latchkernel constructed 3 DummyUnifiedMemoryResource instances (via
make_scratch_buffer) plus a LegacyPinnedMemoryResource (via LatchKernel)
without init_cuda. On Windows and aarch64 the ctypes-pointer/Buffer/MR
chains form cycles that survive refcounting, so the live-MemoryResource
count stays elevated past the test and the next init_cuda test
(test_oom_diagnostics_probe_basics_is_live_and_cheap) is blamed by the
leftover check (UnreleasedMemoryResourceWarning, escalated to an error).

Migrating to init_cuda lets the teardown gate collect the cyclic leftovers,
matching the pattern from 68f832c for the other MR-constructing tests.

Author: juenglin

cuda_core/tests/test_helpers.py

@@ -33,12 +33,11 @@
 
 @pytest.mark.skipif(Device().compute_capability.major < 7, reason="__nanosleep is only available starting Volta (sm70)")
 @pytest.mark.thread_unsafe(reason="requires a barrier wait to avoid overlapping pinned latch allocations")
-def test_latchkernel():
+def test_latchkernel(init_cuda):
     """Test LatchKernel."""
     log = TimestampedLogger(enabled=ENABLE_LOGGING)
     log("begin")
-    device = Device()
-    device.set_current()
+    device = init_cuda
     stream = device.create_stream()
     target = make_scratch_buffer(device, 0, NBYTES)
     zeros = make_scratch_buffer(device, 0, NBYTES)