[Simon Gallagher (vinfnet)]

Summary

Add a self-contained confidential H100 AKS worker sample following Azure's published onboarding guide, and signpost it from the repository and AKS READMEs.

• Use the latest offered AKS Ubuntu node image and require an Azure-FDE kernel at least 6.8.0-1071-azure-fde.
• Install through the guide's GPU Operator path, verify confidential configuration, run strict GPU attestation before and after a TensorFlow workload, and require actual GPU placement rather than CPU fallback.
• Include preflight checks, isolated local state, bounded readiness waits, automatic compute shutdown, offline regression tests, and anonymized attestation examples.
• Document version variability, the distinction between GPU and whole-VM attestation, and the vendor-support boundary for newer AKS images.
• Add prominent sample-discovery links and clarify that automatic shutdown does not delete retained resources.

Validation

• 105 offline regression assertions passed.
• Live West Europe deployment revalidated successfully: strict pre/post-workload GPU attestation and TensorFlow GPU:0 matrix multiplication passed.
• Automatic shutdown verified; all resource groups created for revalidation were subsequently deleted and their absence checked.
• README link targets checked and git diff --check passed.

Private deployment evidence, credentials, kubeconfigs, and local configuration are excluded.