support for compiling cuda using clang

[volatileAntiEntropy]

I'm wondering if there is any equivalency to add_cucodegens("native") for the clang tool chain, or more generally, is there any possibility that we can have a tool chain that uses clang and one pass compilation for cuda code?

[waruqi]

not supported yet, but there a workaround solution. #2305

[volatileAntiEntropy]

Well, I get warnings like warning: add_cuflags("--cuda-gpu-arch=sm_86") is ignored, please pass {force = true} or call set_policy("check.auto_ignore_flags", false) if you want to set it.. However, if I do set_policy("check.auto_ignore_flags", false), nothing changes, I still get that warning.