Issue with cuda::atomic_ref::notify_one and wait

[gevtushenko]

I can see a compilation error for the following code:

__global__ void example_kernel(int* flag_mem) {
  cuda::atomic_ref<int, cuda::thread_scope_device> flag(*flag_mem);
  flag.notify_one();
  flag.wait(0, cuda::std::memory_order_relaxed);
}

The error seems to be reproduceable on CTK 12.0.1 version of the libcu++ as well as on main:

error: no instance of function template "cuda::std::__4::__cxx_atomic_wait" matches the argument list
no instance of function template "cuda::std::__4::__cxx_atomic_notify_one" matches the argument list

[wmaxey]

This should be resolved with #379.