test(bindings): use a capped mempool in test_interop_memPool - #3011

Description

follow-on to #2381

test_interop_memPool in the cuda_bindings Cython and Python tests used the device's default memory pool. Creating the default pool reserves virtual address space of roughly twice the installed device memory, and that reservation is not returned for the lifetime of the process. On address-space-constrained systems this can fail with CUDA_ERROR_OUT_OF_MEMORY even when plenty of device memory is free, which is the failure mode described in #2381. It also follows the convention in cuda_core/tests/AGENTS.md: tests that need a pool should create a capped one.

This PR changes both tests to:

• create a small pool (maxSize = 2 MiB) with cuMemPoolCreate
• set it with cudaDeviceSetMemPool
• read it back with cudaDeviceGetMemPool and assert it is the same handle, then pass it to cuDeviceSetMemPool
• destroy the pool before destroying the context

The capped pool is set before it is queried because cuDeviceGetMemPool and cudaDeviceGetMemPool also create the default pool when no pool has been set. Running the test no longer grows the process virtual size (5 GiB before and after, measured locally on an RTX 6000 Ada).

Both tests still check the driver/runtime handle round trip. The Python test no longer needs the xfail_if_mempool_oom guard. The two *GetDefaultMemPool calls are no longer made in either test, so cuda_bindings no longer exercises them directly.

Checklist

• New or existing tests cover these changes.
• The documentation is up to date with these changes.