test(cuda.core): split managed-memory ops tests into tests/memory/

Move the managed-memory advise/prefetch/discard/discard_prefetch tests
(plus their TestLocation/TestLocationCoerce/TestPrefetch/TestDiscard/
TestDiscardPrefetch/TestAdvise classes and skip helpers) from
test_memory.py into tests/memory/test_managed_ops.py per Andy's nit.

Promote DummyDeviceMemoryResource to helpers.buffers so both files can
import it; the remaining DummyHost/DummyPinned/NullMemoryResource stay
in test_memory.py since they're only used there.

Broader memory-tests reorg ("and siblings": buffer/managed_resource/
pinned/vmm) tracked as a follow-up cleanup PR to keep this diff focused.

Author: rparolin

cuda_core/tests/helpers/buffers.py

@@ -9,6 +9,7 @@
 from . import libc
 
 __all__ = [
+    "DummyDeviceMemoryResource",
     "DummyUnifiedMemoryResource",
     "PatternGen",
     "TrackingMR",
@@ -18,6 +19,30 @@
 ]
 
 
+class DummyDeviceMemoryResource(MemoryResource):
+    def __init__(self, device):
+        self.device = device
+
+    def allocate(self, size, stream=None) -> Buffer:
+        ptr = handle_return(driver.cuMemAlloc(size))
+        return Buffer.from_handle(ptr=ptr, size=size, mr=self)
+
+    def deallocate(self, ptr, size, stream=None):
+        handle_return(driver.cuMemFree(ptr))
+
+    @property
+    def is_device_accessible(self) -> bool:
+        return True
+
+    @property
+    def is_host_accessible(self) -> bool:
+        return False
+
+    @property
+    def device_id(self) -> int:
+        return 0
+
+
 class DummyUnifiedMemoryResource(MemoryResource):
     def __init__(self, device):
         self.device = device