[clang] [CUDA] Add device-side kernel launch support (PR #165519)

[via cfe-commits]

darkbuck wrote:

> Any chances of testing something like this at runtime? I know we have the external test suite but I don't know how easily we could add a trivial test for this stuff

The codegen test is added in this change. For the real test running GPU, I ported dynamic mandelbrot here git at github.com:darkbuck/mandelbrot-dyn.git and verified on GTX 4060.

https://github.com/llvm/llvm-project/pull/165519