add min-cuda-version support for pod create

[vavo]

E-253: add min-cuda-version support for pod create

• resolves Support --min-cuda-version flag for pod create #253 by adding --min-cuda-version to runpodctl pod create for gpu pod placement.
• maintains full legacy command compatibility (stderr warnings only, stdout unchanged).
• keeps runpodctl as the primary binary; drop-in upgrade for existing users.
• aligns the graphql create path with the current RunPod schema, which includes minCudaVersion on GpuAvailabilityInput and PodFindAndDeployOnDemandInput.
• adds unit coverage for command wiring, gpu-only validation, request building, and graphql serialization.
• updates the generated pod create docs.

reference:

• https://graphql-spec.runpod.io/#definition-GpuAvailabilityInput

Test plan:

• go test ./...
• go test -tags e2e ./e2e

[vavo]

live verification completed with a real Runpod API key.

verified flow:

• build runpodctl
• gpu list confirmed available gpu ids, including NVIDIA GeForce RTX 4090
• create disposable pod with:
  • --image ubuntu:22.04
  • --gpu-id 'NVIDIA GeForce RTX 4090'
  • --min-cuda-version 12.6
  • --name codex-issue-253-probe
• create succeeded and returned pod id nuyi29y19kylna
• immediately deleted the disposable pod with runpodctl pod delete nuyi29y19kylna

result: pass

cleanup completed successfully.

[promptless]

Promptless prepared a documentation update related to this change.

Triggered by PR #265

Added the --min-cuda-version flag to the runpodctl pod create command reference with a usage example and description noting it applies only to GPU Pods.

Review at https://app.gopromptless.ai/suggestions/c495ba32-59bf-4ed6-9546-20a09ba1f689