More reliable macOS CI fleet on Scaleway Apple Silicon (auto-recovery from stuck nodes and wedge-prone processors)

Improves fleet resilience for macOS CI hosts running on Scaleway Apple Silicon. Nodes that previously got stranded in Ready=Unknown after a kubelet heartbeat gap (for example during a fleet-wide restart) now recover automatically and become schedulable again, and xcresult processor VMs that could silently wedge when the Tailscale/NAT path failed now self-recover within about a minute instead of needing a manual recycle. Users running builds on Tuist’s Scaleway Apple Silicon CI should see fewer stuck runs, fewer orphaned jobs, and faster recovery after network blips affecting the macOS fleet.