[llvm] [Offload] Add olGetKernelMaxGroupSize (PR #142950)

[Joseph Huber via llvm-commits]

jhuber6 wrote:

We already have an internal kernel type, don't we? Likely we just need something similar to the set / get attributes for kernels that CUDA / HSA provide. Though I suppose that this is just another level because most would probably be at the device level? Like I said, HSA already does this so there's some precedent.

https://github.com/llvm/llvm-project/pull/142950