1. Proposed title of this feature request

Ability to set quota for nvidia.com/gpu resource on namespace level

2. What is the nature and description of the request?

Limiting use of nvidia.com/gpu resource on bare metal nodes with nVidia GPUs is not possible at the moment. 

Customer needs a way to restrict or deny nvidia.com/gpu usage by a specific namespace.

3. Why does the customer need this? (List the business requirements here)

Workload/Deployment in any namespace is automatically claiming nvidia.com/gpu resource, even if it does not need it. This leads to lack of nvidia.com/gpu resource for the deployments in specific namespaces for which using GPUs is mandatory.

4. List any affected packages or components.

nVidia Operator, Scheduler, Workload