Add CI for GPU tests on Ruche #3

Job	Run time
check_docker_files	5s
lint	9s
build_base / build_base (openmp, false)	1m 52s
build_base / build_base (cuda, true)	9m 29s
build_base / build_base (hip, false)	4m 9s
build_base / build_base (sycl, false)	3m 29s
build (openmp, gcc, g++, -DKokkos_ENABLE_OPENMP=ON, native)	4m 9s
build (cuda, gcc, g++, -DKokkos_ENABLE_CUDA=ON -DKokkos_ARCH_AMPERE80=ON, native)	7m 20s
build (cuda, gcc, g++, -DKokkos_ENABLE_CUDA=ON -DKokkos_ARCH_AMPERE80=ON, host_device, -DKokkosFF...	7m 6s
build (hip, hipcc, hipcc, -DKokkos_ENABLE_HIP=ON -DKokkos_ARCH_VEGA90A=ON, native)	15m 20s
build (hip, hipcc, hipcc, -DKokkos_ENABLE_HIP=ON -DKokkos_ARCH_VEGA90A=ON, host_device, -DKokkosF...	12m 42s
build (sycl, icx, icpx, -DKokkos_ENABLE_SYCL=ON -DKokkos_ARCH_INTEL_GEN=ON, native)	10m 23s
build (sycl, icx, icpx, -DKokkos_ENABLE_SYCL=ON -DKokkos_ARCH_INTEL_GEN=ON, host_device, -DKokkos...	9m 38s
test (cuda, self-hosted, cuda)	41m 14s
test (openmp, ubuntu-latest)	22s
	2h 7m 27s

Provide feedback