[k8s] Fix GKELabelFormatter for H100s

[romilbhardwaj]

get_gke_accelerator_name returns the incorrect label for H100 (nvidia-tesla-h100 instead of nvidia-h100). This PR fixes it by removing the incorrect conditional check for H100.

Tested:

• sky launch --gpus H100:1 on a GKE cluster with H100s.

[romilbhardwaj]

H100s are a hard to get on GKE, so to test this PR I mocked it on my sky local up cluster with:

kubectl proxy

curl --header "Content-Type: application/json-patch+json" \
  --request PATCH \
  --data '[{"op": "add", "path": "/status/capacity/nvidia.com~1gpu", "value": "8"}]' \
  http://localhost:8001/api/v1/nodes/skypilot-control-plane/status


kubectl label nodes skypilot-control-plane cloud.google.com/gke-accelerator=nvidia-h100-80gb

With that, sky launch --gpus H100:1 works as expected, and the gpu name is now H100 (instead of H100-80GB).