refactor: Restructure manifests/ into Kustomize base/overlay layout

Context

All manifests currently live as flat YAML in manifests/.

• PR #65 is already in review doing this restructure, but there's no tracking issue — so there's nowhere to discuss scope or link downstream work.

Problem

Without a base/overlay structure, every new manifest gets added as a flat file that's not portable across environments. You can't patch for dev vs. prod without duplicating files.

What This Issue Tracks

Restructure manifests/ into:

• base/ — portable resources any K8s cluster can apply
• overlays/oci/ — OCI/OKE-specific patches
• overlays/generic/ — local dev (Minikube/kind)

Files to move: serving-runtime.yaml, inference-service.yaml, milvus-deployment.yaml (strip the status: block), and the deployment/service YAMLs under server-https/.

Why This Order

#47 (scale-to-zero), #48 (KServe InferenceService), the OCI deployment work, and the CORS fix (#76) all require overlays to exist before their patches can land. Nothing in the manifest layer moves cleanly without this structure first.

Acceptance Criteria

• kubectl apply -k manifests/overlays/generic/ works on a local cluster
• kubectl apply -k manifests/overlays/oci/ applies OCI-specific patches on top
• No flat YAMLs remain directly in manifests/
• PR #65 linked and closed by this issue

Related

• PR #65 — implementation already in review
• Parent tracking issue: #72
• Blocks:
  • #76
  • #47
  • #48 and the OCI deployment work

Start by reviewing PR #65 and the YAML files listed under manifests/, including server-https/. Verify the base and generic/OCI overlay layout, then run kubectl apply -k manifests/overlays/generic/ and kubectl apply -k manifests/overlays/oci/; done means no flat YAMLs remain, the overlays apply, and PR #65 is linked and closed.