feat(backends): Pool-Modell — Backend = Pool, N Server pro Backend

Migration 0016: backend_servers (id, backend_id, name, address, port,
weight, backup, active) + backends.lb_algorithm. Daten-Migration kopiert
bestehende backends.address/port als ersten Server, dann DROP COLUMN.

HAProxy-Renderer: rendert pro Backend einen Block mit `balance <algo>`
+ N `server`-Zeilen (weight, backup-Flag, optional check inter 5s).
LB-Algorithmen: roundrobin / leastconn / source.

REST: /backends/:id/servers (GET/POST), /backend-servers/:id (PUT/DELETE).
Re-rendert HAProxy nach jeder Server-Mutation.

UI: address/port aus Backend-Form raus, lb_algorithm-Select rein. Server
verwaltet ein expandable Sub-Panel pro Backend-Row (Tabelle + Add/Edit/
Delete-Modal). Domain-Attachment-Multi-Select bleibt.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

management-ui/src/i18n/locales/de/common.json

@@ -202,13 +202,11 @@
   },
   "backends": {
     "title": "Backends",
-    "intro": "Upstream-Server, an die HAProxy weiterroutet. Health-Check-Pfad optional aktiviert TCP+HTTP-Probes alle 5s.",
-    "addBackend": "Backend hinzufügen",
-    "editBackend": "Backend bearbeiten",
+    "intro": "Upstream-Pools (Backend = N Server). HAProxy verteilt laut LB-Algorithmus; Health-Check-Pfad aktiviert HTTP-Probes alle 5s pro Server.",
+    "addBackend": "Backend-Pool hinzufügen",
+    "editBackend": "Backend-Pool bearbeiten",
     "name": "Name",
     "scheme": "Schema",
-    "address": "Adresse",
-    "port": "Port",
     "target": "Ziel",
     "healthCheck": "Health-Check-Pfad",
     "active": "Aktiv",
@@ -217,8 +215,30 @@
     "attachedDomains": "Domains",
     "attachedDomainsHint": "Domains, die dieses Backend als Primary verwenden. Auswahl umkonfiguriert die Domains direkt — gleiche Quelle wie der Backend-Picker im Domain-Modal.",
     "selectDomains": "Domains wählen",
+    "lbAlgo": "Load-Balancing",
+    "lbAlgoHint": "roundrobin = gleichmäßig, leastconn = an den Server mit wenigsten Verbindungen, source = sticky per Client-IP (für stateful Apps ohne shared session).",
+    "servers": "Server",
+    "noServers": "kein Server",
+    "nServers": "{{n}} Server",
+    "serversIn": "Server in „{{name}}\"",
+    "serverHintCreate": "Speichern legt nur den Pool an. Server kommen im nächsten Schritt — Pool öffnen → „Server hinzufügen\".",
     "actions": "Aktionen",
-    "deleteConfirm": "Backend {{name}} wirklich löschen?"
+    "deleteConfirm": "Backend-Pool {{name}} wirklich löschen? Alle Server im Pool werden mitentfernt.",
+    "server": {
+      "intro": "Upstream-Server im Pool. Reihenfolge in HAProxy egal — der LB-Algorithmus entscheidet.",
+      "add": "Server hinzufügen",
+      "edit": "Server bearbeiten",
+      "name": "Server-Name",
+      "address": "Adresse",
+      "port": "Port",
+      "target": "Endpoint",
+      "weight": "Gewicht",
+      "weightHint": "0–256. Höher = mehr Traffic. 100 = Standard.",
+      "backup": "Backup",
+      "backupHint": "Backup-Server werden nur angesprochen, wenn alle primären Server (non-backup) down sind.",
+      "empty": "Noch keine Server im Pool. „Server hinzufügen\" startet damit.",
+      "deleteConfirm": "Server {{name}} wirklich löschen?"
+    }
   },
   "routing": {
     "title": "Routing-Regeln",

management-ui/src/i18n/locales/en/common.json

@@ -202,13 +202,11 @@
   },
   "backends": {
     "title": "Backends",
-    "intro": "Upstream servers HAProxy proxies to. Optional health-check path enables TCP + HTTP probes every 5s.",
-    "addBackend": "Add backend",
-    "editBackend": "Edit backend",
+    "intro": "Upstream pools (one backend = N servers). HAProxy balances load by the chosen algorithm; health-check path enables HTTP probes every 5s per server.",
+    "addBackend": "Add backend pool",
+    "editBackend": "Edit backend pool",
     "name": "Name",
     "scheme": "Scheme",
-    "address": "Address",
-    "port": "Port",
     "target": "Target",
     "healthCheck": "Health check path",
     "active": "Active",
@@ -217,8 +215,30 @@
     "attachedDomains": "Domains",
     "attachedDomainsHint": "Domains that use this backend as their primary. Selecting domains here reconfigures them directly — same source of truth as the Domain modal's backend picker.",
     "selectDomains": "Select domains",
+    "lbAlgo": "Load balancing",
+    "lbAlgoHint": "roundrobin = evenly, leastconn = pick the server with fewest active connections, source = sticky per client-IP hash (for stateful apps without shared session).",
+    "servers": "Servers",
+    "noServers": "no server",
+    "nServers": "{{n}} servers",
+    "serversIn": "Servers in “{{name}}”",
+    "serverHintCreate": "Saving creates the pool only. Add servers in the next step — open the pool and click “Add server”.",
     "actions": "Actions",
-    "deleteConfirm": "Really delete backend {{name}}?"
+    "deleteConfirm": "Really delete backend pool {{name}}? All servers in the pool will be removed too.",
+    "server": {
+      "intro": "Upstream servers in this pool. Order doesn't matter — the LB algorithm decides.",
+      "add": "Add server",
+      "edit": "Edit server",
+      "name": "Server name",
+      "address": "Address",
+      "port": "Port",
+      "target": "Endpoint",
+      "weight": "Weight",
+      "weightHint": "0–256. Higher = more traffic. 100 = default.",
+      "backup": "Backup",
+      "backupHint": "Backup servers receive traffic only when every primary (non-backup) server is down.",
+      "empty": "No servers in the pool yet. Click “Add server” to get started.",
+      "deleteConfirm": "Really delete server {{name}}?"
+    }
   },
   "routing": {
     "title": "Routing rules",