Recast KEP core model to a spoke-reconciled lifecycle

[roguepikachu]

Summary

Recasts the Cluster Infrastructure KEP core model so the spoke is self-sufficient:

• The hub owns a SpokeCluster per managed cluster: the fleet handle you list with kubectl get spokeclusters and the dispatcher that hands a blueprint revision to the spoke.
• The spoke runs vela-cluster-core, which reconciles the dispatched ClusterBlueprint into a self-reconciling Cluster and keeps it converged locally, with or without the hub.
• Status push-back is removed. The hub reads spoke state by querying on demand, so hub downtime never stops a spoke from reconciling.
• The lifecycle phases are reframed to infraProvisioning (hub), planeProvisioning (spoke), and healthValidation (hub/both).

Sections changed

• The Approach; CRDs Introduced (adds SpokeCluster, recasts Cluster)
• What Lives Where (hub vs spoke responsibilities, single-pane-of-glass callout)
• Relationship to Existing Multi-Cluster Architecture: Proposed Architecture, Design Principles, Controller Ownership Model, Migration Path, Controller Reconciliation, Controller Responsibilities Matrix
• Core CRDs: adds a SpokeCluster summary, recasts the Cluster intro
• Cluster Lifecycle Phases (renamed and rewritten)
• Table of Contents anchors

Deferred (follow-up)

The downstream sections still describe the prior hub-centric flow and are not yet updated: the Cluster spec catalog (some fields, for example credential and rolloutStrategyRef, move to SpokeCluster), the provision/adopt CLI, rollouts, health, and the Use Cases. This is called out inline with a field-migration note on the Cluster CRD.

Test plan

• Markdown renders and the TOC links resolve for the renamed and added headings
• Review the model for consistency with the Phase 1 connect plan (GWCP-100435)