Worksizes should depend on specific Kernel (Feature #172)
In the end, local- and global worksizes (ls, gs) should depend on the specific kernel.
At the moment, each kernel is called within a function ending on ..._device using the general function enqueueKernel. Before this happens, the function
is called (the "2" is just because there is an older function still in use at some points) to get ls, gs.
To get_work_sizes2, also the kernel is passed as an argument, which gives the opportunity within the function to query for specific ls, gs.