Shared GPU (`nvidia.com/gpu.shared`) is shown as `n/a` on K9s node view

[hoangcongchuc]

Is your feature request related to a problem? Please describe.
Yes. K9s currently does not display the number of shared GPUs (nvidia.com/gpu.shared) on Kubernetes nodes. Nodes using shared GPU resources show n/a in the GPU column on the Nodes view. This leads to confusion, as users might assume those nodes have no GPU capacity available, when in fact they do.

Describe the solution you'd like
I would like K9s to support displaying the value of nvidia.com/gpu.shared in addition to nvidia.com/gpu. Ideally:

If nvidia.com/gpu.shared is present, it should be shown with its value.

If both gpu and gpu.shared exist, both should be listed or made clear.

This would greatly help in monitoring heterogeneous clusters that use both full and shared GPU allocations.

Describe alternatives you've considered

Checking GPU availability manually via kubectl describe node, which is time-consuming.

Using third-party monitoring dashboards like Grafana + DCGM exporter, which introduces more operational overhead.

Additional context
Example output from kubectl describe node:

Capacity:
nvidia.com/gpu: 0
nvidia.com/gpu.shared: 24

In this case, K9s shows n/a, but the node is GPU-capable using shared GPUs.

This feature would be valuable for any teams using NVIDIA GPU Operator with shared GPU scheduling in Kubernetes.