Investigate supporting GKE autopilot and EKS Fargate

[michaeljguarino]

Use Case

I suspect the pricing models of both of these might outperform a highly overprovisioned minimal cluster, so we should double-check this and give users an option to enable this for their use cases. There are a few things we should verify though:

[] stateful workloads still work (postgres esp)
[] cost assumptions are correct
[] if there's a migration path to non-serverless k8s

Ideas of Implementation

create a new test cluster with those settings, hammer on it

---

Message from the maintainers:

Excited about this feature? Give it a 👍. We factor engagement into prioritization.