Instant VM/LXC modals via bulk hydration + server prewarmer

Backend keeps every guest's modal payload (details, backups, apps, schedule, mount points) warm in-memory and exposes them through a single `/api/vms/modal-cache-all` endpoint. The dashboard hydrates its entire modal cache from that one request on page load, so opening any guest — first click or after coming back later — renders instantly. Replaces ~84 per-guest fetches with 1.

AppImage/lib/lxc-apps-cache.ts

@@ -0,0 +1,58 @@
+// Shared cross-component cache for the LXC App-tab payload
+// (registered sidecar + auto-detected suggestions). Used by both
+// virtual-machines.tsx (which prefetches on modal open and on hover)
+// and lxc-app-panel.tsx (which reads the cache first and only fetches
+// if empty). The in-flight promise map dedups concurrent requests: if
+// the parent already fired a prefetch, the panel awaits the SAME
+// promise instead of duplicating the request against the backend —
+// no more racing fetches on tab switch during a slow first visit.
+
+import { fetchApi } from "./api-config"
+
+export type LxcAppsBundle = {
+  sidecar: any
+  suggestions: any | null
+}
+
+const dataCache = new Map<number, LxcAppsBundle>()
+const inFlight = new Map<number, Promise<LxcAppsBundle | null>>()
+
+export function getLxcAppsCached(vmid: number): LxcAppsBundle | undefined {
+  return dataCache.get(vmid)
+}
+
+export function fetchLxcApps(vmid: number): Promise<LxcAppsBundle | null> {
+  const existing = inFlight.get(vmid)
+  if (existing) return existing
+  const p = Promise.all([
+    fetchApi(`/api/vms/${vmid}/apps`).catch(() => null) as Promise<any>,
+    fetchApi(`/api/vms/${vmid}/apps/suggestions`).catch(() => null) as Promise<any>,
+  ])
+    .then(([sc, sug]) => {
+      if (!sc) return null
+      const bundle: LxcAppsBundle = { sidecar: sc, suggestions: sug }
+      dataCache.set(vmid, bundle)
+      return bundle
+    })
+    .finally(() => {
+      inFlight.delete(vmid)
+    })
+  inFlight.set(vmid, p)
+  return p
+}
+
+export function invalidateLxcApps(vmid: number): void {
+  dataCache.delete(vmid)
+}
+
+// Seed the cache with a sidecar payload from the bulk modal-cache
+// endpoint. Only the sidecar side is populated — suggestions still
+// resolve lazily when the App panel actually mounts (the bulk
+// endpoint intentionally excludes them since most guests don't
+// need the auto-detected chips and the payload would balloon).
+export function seedLxcAppsCache(vmid: number, sidecar: any): void {
+  if (!sidecar) return
+  const existing = dataCache.get(vmid)
+  if (existing) return  // per-panel fetch already ran, don't overwrite
+  dataCache.set(vmid, { sidecar, suggestions: null })
+}