[llvm] [Offload] Store kernel name in GenericKernelTy (PR #142799)

[Ross Brunton via llvm-commits]

RossBrunton wrote:

@jhuber6 Looking at this again, as far as I can tell this field is used in at least the following places:

* When building an AMDGPU kernel.
* When generating an error when launching kernels in the CUDA and AMDGPU adapters.
* When printing an error from a kernel trap.
* Debug messages when launching a kernel.
* Used in the Record/Replay interface to identify kernels.

Each of these locations could create their own copy of the kernel name or change their API to allow passing it in, but I'm not sure the benefit of that over just having a copy of the name stored in `GenericKernelTy`.

https://github.com/llvm/llvm-project/pull/142799