fix profiling profiling_timebase

[rjodinchr]

Use explicit work_group_size of 1

Fix #2240

[bashbaug]

I'm OK with this change since I don't think it's an important part of the test to have a global_work_size equal to NULL, but I do believe this part of the test is correct, and a global_work_size equal to NULL is valid - at least for OpenCL 2.1 and newer. I've filed a few issues to get this clarified, and to ensure that the case where the global_work_size is equal to NULL is tested:

KhronosGroup/OpenCL-Docs#1302
#2244

Note, clGetDeviceAndHostTimer also requires OpenCL 2.1, and there's no check for OpenCL 2.1 or newer (that I can see at least), so this really ought to be added. If you'd prefer to add this check in a separate PR, would you mind filing an issue so it doesn't slip through the cracks? I suspect it will be less work to just add the check, though 😄

[rjodinchr]

Alright, I missed that part of the spec. Honestly having something about global_work_size being NULL not in the global_work_size section but the work_dim section is really easy to miss.

In that case, I think we should not land that PR. I'll fix clvk according to the spec.