feat: add scalable multi-node kubeadm quickstart

[convee]

Summary

• make the standalone kubeadm quickstart a real three-node cluster by default: one tainted control plane plus two workers
• add safe worker scaling from 1 to 4 nodes with drain, local-PV affinity protection, image import, rollout, and public HTTP re-verification
• keep the NodePort localhost URLs working when workloads move between workers, including after control-plane restart
• document topology, resource requirements, scaling commands, safety boundaries, and the non-HA scope in README and deployment guides
• update the GitHub Pages homepage with the multi-node and scale-out story

Validation

• 1031 tests passed, 1 skipped
• console lint, typecheck, and production build passed
• Helm lint/render, homepage contract, Lima validation, shell syntax, and generated diff checks passed
• anonymous fresh-clone install reached 3/3 Ready nodes and deployed hello-site on a worker
• scaled 2 → 3 → 1 → 2 workers, with drain and public body-digest verification at each step
• rescheduled the application from worker 1 to worker 2 and verified the published localhost URL still returns HTTP 200
• restarted the control-plane VM and verified the relay timer restored the same HTTP 200 route
• browser E2E verified first-login guidance, dashboard data, the Applications/Monitoring/Merchants/Tenants/Builds modules, and the application Open link

Boundary

This is functional multi-node placement and worker elasticity for local evaluation. It is intentionally not a highly available control plane.