feat: add Oracle OKE recipe overlays

[mchmarny]

Summary

As part of the ongoing effort to expand AICR recipe coverage across CSP and NCP managed Kubernetes offerings, we should add support for Oracle Kubernetes Engine (OKE). AICR already supports AKS (Azure), EKS (AWS), and GKE (Google Cloud) — OKE is the next logical addition to round out coverage of major cloud providers.

Scope

• Add base oke.yaml overlay with OKE-specific service criteria and defaults
• Add intent overlays: oke-training.yaml, oke-inference.yaml
• Add accelerator-specific overlays for supported GPU types (e.g., h100-oke-training.yaml, h100-oke-inference.yaml)
• Add OS/framework variant overlays as needed (e.g., h100-oke-ubuntu-training.yaml)
• Add OKE-specific component values if GPU Operator or other components require OKE-specific configuration
• Add KWOK node simulation scripts for OKE node topologies
• Validate all new overlays via KWOK tests

Context

This is part of a broader effort to ensure AICR provides validated GPU-accelerated configurations across all major managed Kubernetes platforms:

Provider	Service	Status
AWS	EKS	Supported
Azure	AKS	Supported
Google Cloud	GKE	Supported
Oracle Cloud	OKE	This issue

Notes

• Follow existing overlay patterns established by EKS/AKS/GKE
• Reference ADR-003 (docs: add ADR-003 for scaling KWOK recipe tests #424) for CI scaling considerations when adding new overlays
• OKE-specific GPU node labels, taints, and topology should be captured in KWOK simulation scripts

[OguzPastirmaci]

@mchmarny I'm one of the maintainers of the OKE GPU deployment stack and would be happy to help.

[mchmarny]

Excellent, welcome to the project, @OguzPastirmaci. Couple of good docs to tarts with:

• Recipe Development Guide
• Data Architecture

If helpful we can also do a quick call. Few other maintainers here that can help: @yuanchen8911 @dims. You can find all of us in the k8s Slack as well. https://kubernetes.slack.com/archives/C09N46EFJR0

Looking forward to collaborating on this