Add fallback memory resource for TCC devices

[ksimpson-work]

For devices which don't support memory pools, we need to provide an alternate default memory resource.

This basic WAR implementation works. I used a colossus lease for a Tesla T4 on Friday Nov 29 and these were the results:

using the DefaultAsyncMempool --> python - m pytest tests/test_memory.py
=============================================== short test summary info ===============================================
FAILED tests/test_memory.py::test_buffer_initialization - cuda.core.experimental._utils.CUDAError: CUDA_ERROR_NOT_SUPPORTED: operation not supported

using the implementation in this branch --> python - m pytest tests/test_memory.py
collected 4 items

tests\test_memory.py .... (SUCCESS)

close #208

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[ksimpson-work]

re-tested using synchronous malloc and free on Tesla T4 colossus instance.

on main branch:

(test_env) C:\cuda-python\cuda_core>python -m pytest tests\test_memory.py

platform win32 -- Python 3.12.7, pytest-8.3.4, pluggy-1.5.0
rootdir: C:\cuda-python\cuda_core
configfile: pyproject.toml
collected 4 items

tests\test_memory.py FFFF [100%]

with change:

(test_env) C:\cuda-python\cuda_core>python -m pytest tests\test_memory.py

platform win32 -- Python 3.12.7, pytest-8.3.4, pluggy-1.5.0
rootdir: C:\cuda-python\cuda_core
configfile: pyproject.toml
collected 4 items

tests\test_memory.py .... [100%]

[leofang]

/ok to test

[leofang]

Windows failures are known (#271) and irrelevant. Let's merge. Thanks, Keenan!