[release/2.9] Fix test OOMs triggered by test_cuda.py (#2816)

Cherry-pick of
pytorch@e4ffd71

> This PR fixes a bug in the `test_garbage_collect_expandable` unit test
where the finally block incorrectly re-reads the current per process
memory fraction instead of setting the original value. With out the fix
the other tests in the `test/test_cuda.py` test suite were impacted and
failed with OOM error on ROCm.
> 
> This ensures proper cleanup and isolation of test state, maintaining
test correctness and avoiding side effects like the below OOM error that
it caused.

Progress on ROCm/TheRock#2118

Co-authored-by: rraminen <[email protected]>

Author: ScottTodd

test/test_cuda.py

@@ -4408,7 +4408,7 @@ def alloc(n):
             # expandable_segment blocks can be in the free list when this is called.
             alloc(80)
         finally:
-            orig = torch.cuda.get_per_process_memory_fraction(0)
+            torch.cuda.memory.set_per_process_memory_fraction(orig)
 
     def test_allocator_settings(self):
         def power2_div(size, div_factor):