Prevent macOS CI builds from hanging due to lost cache node private network address

Fixed an issue where runner-cache Elastic Metal nodes on Scaleway could silently lose their Private Network (pn0) IP address when DHCP renewals stopped being acknowledged. Because the node still appeared healthy on its public network, the failure was invisible to the control plane, but every macOS runner’s request to the module cache was blackholed. macOS CI builds then hung at the remote module-cache fetch step with long timeouts per module. The node now discovers the Private Network address via DHCP once and holds it statically, so the cache remains reachable across reboots, lease churn, and silent DHCP servers.