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

[Ross Brunton via llvm-commits]

RossBrunton wrote:

@jhuber6 This has been sitting here for a while now, and I'd like to either merge it or close it. I think practically speaking it has to be it's own function different from the getKernelInfo function because of the dynamic size requirement.

I think this makes sense to add to the liboffload API. There's an equavilent API in cuda which seems to be commonly used, and providing it in liboffload (and for AMD when it gets implemented) will make porting code easier.

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