Runner-cache builds now fall back to the public cache when a private Kura node is unready

Fixed a bug where builds were routed to a dead private cache endpoint when the underlying Kura node went unready, causing module-cache downloads to time out. The runner-cache dispatch is now readiness-aware: when a private node-port server stops reporting ready, builds are automatically degraded to the public cache instead of timing out. This prevents widespread CI and macOS runner build failures during Kura mesh outages.