As a RHODS user, I want to be able to select GPUs for my notebook server environment, so I can accelerate model development activities.

Requirements:
1. P0: The system must provide an option to select CUDA notebook image.
2. P0: When creating notebook server environments, the system must present the number of GPUs to select based on what's available in the cluster. If there are no GPUs available, the option must be hidden.
3. P0: Only notebook server sessions with GPUs selected should utilize GPUs. The goal is to ensure workloads that don't need GPUs do not unnecessarily utilize GPUs, so GPUs are reserved for workloads that need them.
4. P0: Customers that purchase RHODS must be entitled to install the GPU operator.