SWDEV-257789 - Initial change to skip kernel arg copy

The optimization is controlled with ROCR_SKIP_KERNEL_ARG_COPY.
This is initial check-in for experiments. Extra changes are
necessary for full support:
- handle graph capture with the original sysmem alloc
- avoid memobject references, otherwise there is a race condition with
reusage of the arg buffer
- Remove arg setup from hip

Change-Id: Ib0af710f93e79834711fa4049a7c66093711e68b


[ROCm/clr commit: 7e12cf6318]

projects/clr/rocclr/platform/command.cpp

@@ -633,7 +633,8 @@ int32_t NDRangeKernelCommand::captureAndValidate() {
 
   int32_t error;
   uint64_t lclMemSize = kernel().getDeviceKernel(device)->workGroupInfo()->localMemSize_;
-  parameters_ = kernel().parameters().capture(device, sharedMemBytes_ + lclMemSize, &error);
+  parameters_ = kernel().parameters().capture(*queue()->vdev(),
+                                              sharedMemBytes_ + lclMemSize, &error);
   return error;
 }