fix: right-size service memory after PVE RAM upgrade (142→272GB)

- MySQL InnoDB: 2Gi/4Gi → 3Gi/6Gi (was at 97% of limit)
- Redis HAProxy: 16Mi/16Mi → 32Mi/64Mi (OOMKilled)
- Plotting-book: 64Mi/64Mi → 128Mi/256Mi (OOMKilled)
- Tandoor: 256Mi/256Mi → 384Mi/512Mi (60 OOM restarts), re-enabled
- Navidrome: 128Mi/128Mi → 256Mi/384Mi
- Matrix: add explicit 256Mi/512Mi resources
- Trading-bot workers: 64Mi/64Mi → 128Mi/256Mi, re-enabled
- Tier 3-edge defaults: 96Mi/192Mi → 128Mi/256Mi
- Fallback tier defaults: 128Mi/128Mi → 128Mi/192Mi, max 2→4Gi
- Mailserver: disable rspamd-redis, fix Roundcube IPv6/IMAP, bump dovecot connections

stacks/matrix/main.tf

@@ -115,6 +115,15 @@ resource "kubernetes_deployment" "matrix" {
             name       = "extra-packages"
             mount_path = "/extra-packages"
           }
+          resources {
+            requests = {
+              cpu    = "25m"
+              memory = "256Mi"
+            }
+            limits = {
+              memory = "512Mi"
+            }
+          }
         }
         volume {
           name = "data"