added runtimeClassName to fix Cuda version error on gpu-pod.yaml test

[armagankaratosun]

The example will fail (as it is also reported in the issue https://github.com/NVIDIA/gpu-operator/issues/415)

Failed to allocate device vector A (error code CUDA driver version is insufficient for CUDA runtime version)!
[Vector addition of 50000 elements]

This is not related to the CUDA driver version or runtime on the node but the default runtime class is not set to the nvidia.

Setting the runtimeClassName statically will fix the issue and it will be more straightforward for a new user to test the GPU operator deployment.