Test custom AMI for EFA support

[mikkeloscar]

This PR is purely for enabling testing of an in-progress AMI to enable support for EFA on g5 and p5 instances.

The idea is that a cluster switched to this PR branch will be able to run a node-pool like this:

- config_items:
    labels: dedicated=gpu-g5-efa,zalando.org/nvidia-gpu=nvidia-a10g,efa=enabled
    taints: dedicated=gpu-g5-efa:NoSchedule
    kuberuntu_distro_worker: "jammy_with_gpu"
    tag_instance_infrastructure_component: "false"
    internal_node_subnets_enabled: "true"
  discount_strategy: none
  instance_types:
  - g5.8xlarge
  - g5.12xlarge
  - g5.16xlarge
  - g5.24xlarge
  - g5.48xlarge
  max_size: 30
  min_size: 0
  name: gpu-g5-efa
  profile: worker-karpenter

With will run the special AMI defined as: kuberuntu_image_v1_31_jammy_with_gpu_amd64. To change the AMI to test, only that config-item needs to be changed in this PR and the node pool will be updated.

[mikkeloscar]

@desinurch FYI: I merged latest beta branch into this one to keep it up to date.

[mikkeloscar]

@desinurch This PR is not in a mergable state, just enabled for testing.

We need to update the main AMI which we can't do without rolling the nodes, so it must be either done e.g. with Kubernetes v1.32 (or other changes) or we split out the GPU AMI as we discussed and then we can make a PR just for GPU AMI update.