SWDEV-363074 - Clean-up sync between SDMA and compute

HIP can't rely on the resource tracking, used in OCL and requires different explicit sync.
Make sure ROCCLR syncs compute only when SDMA is used and vise versa.
The new logic will allow to enable CPDMA without unnecessary waits.

Change-Id: Ib9d1788cfd5afa5ea2fec4c96a37d8b9c4d0059d


[ROCm/clr commit: ff6b4db70b]

projects/clr/rocclr/device/pal/palresource.cpp

@@ -1430,6 +1430,7 @@ bool Resource::partialMemCopyTo(VirtualGPU& gpu, const amd::Coord3D& srcOrigin,
     // Make sure compute is done before CP DMA start
     gpu.addBarrier(RgpSqqtBarrierReason::MemDependency);
   } else {
+    gpu.releaseGpuMemoryFence();
     gpu.engineID_ = SdmaEngine;
   }