[UR] Add new urKernelGetSuggestedLocalWorkSizeWithArgs() API (#21442)
We want to remove the possibility of setting arguments
separately from launching a kernel. However, the OpenCL spec
requires all kernel arguments to be set before calling
`urKernelGetSuggestedLocalWorkSize()`,
so we have to add a new
`urKernelGetSuggestedLocalWorkSizeWithArgs()` API
in order to make our goal possible.
`urKernelGetSuggestedLocalWorkSizeWithArgs()` has to call
`urKernelGetSuggestedLocalWorkSize()` only for now,
because its full implementation requires more changes in the sanitizers'
layer,
which will be implemented in the following pull requests.
See the conversation:
https://github.com/intel/llvm/pull/21290#discussion_r2840525187
Ref: #21290
---------
Signed-off-by: Lukasz Dorau <lukasz.dorau@intel.com>