Cluster health remediation: cleanup CronJob, disable Collabora, fix GPU probe, add NFS exports [ci skip]

- Add daily CronJob to auto-clean Failed/Evicted pods cluster-wide (infra-maintenance)
- Disable Collabora in Nextcloud (broken HPA caused scaling storm; using OnlyOffice instead)
- Increase gpu-pod-exporter liveness probe timeout from 1s to 5s
- Add osm-routing NFS exports (osrm-data, otp-data)

modules/kubernetes/nvidia/main.tf

@@ -605,6 +605,7 @@ resource "kubernetes_daemonset" "gpu_pod_exporter" {
             }
             initial_delay_seconds = 30
             period_seconds        = 30
+            timeout_seconds       = 5
           }
         }