[llvm] [Offload] Enable querying a kernel's program (PR #142632)

[Ross Brunton via llvm-commits]

RossBrunton wrote:

@jhuber6 Looking at it again and discussing internally, is this change what you're looking for? Specifically, with extending ol_kernel_t to be a handle type like program, event, etc.

If it's not, and the kernel handle should just stay as a `void *`, I'm not sure what kernel queries I could actually implement. In UR at least ( https://oneapi-src.github.io/unified-runtime/core/api.html#ur-kernel-info-t ), all the properties are either not implementable on Cuda/Hip (arg count) or could be tracked by the offload user (name, context, program).

https://github.com/llvm/llvm-project/pull/142632