[release/2.5] Update CUDAPluggableAllocator.h (#1999)

Altering the flag to use the correct streamType for
CUDAPluggableAllocator. This is impacting Distributed Fused Adam in
Rocm/APEX.

`2025-03-25T14:19:25.0187731Z #5 1815.5 FAILED:
/root/dockerbuild/pytorch.1.7.0a0/apex/build/temp.linux-x86_64-cpython-312/apex/contrib/csrc/nccl_allocator/NCCLAllocator_hip.o`

```
torch/include/torch/csrc/cuda/CUDAPluggableAllocator.h:125:8: error: ‘void torch::cuda::CUDAPluggableAllocator::CUDAPluggableAllocator::recordStream(const c10::DataPtr&, torch::cuda::CUDAPluggableAllocator::streamType)’ marked ‘override’, but does not override
2025-03-25T14:19:25.0192670Z #5 1815.5   125 |   void recordStream(const c10::DataPtr&, streamType stream) override;
2025-03-25T14:19:25.0192737Z #5 1815.5       |        ^~~~~~~~~~~~
```

Apex Commit : ROCm/apex#188

Author: amd-sriram

torch/csrc/cuda/CUDAPluggableAllocator.h

@@ -37,7 +37,7 @@ struct TORCH_CUDA_CPP_API CUDAPluggableAllocatorDeleterContext {
   cudaStream_t stream_;
 };
 
-#if defined(TORCH_HIP_VERSION)
+#if defined(USE_ROCM)
 using streamType = c10::hip::HIPStream;
 #else
 using streamType = c10::cuda::CUDAStream;