perf: fix N+1 query pattern across all modules + rack numbering

CRITICAL PERF BUG: Every hook did storage.list() (1 HTTP call fetching ALL
items with values, discarding values, returning only keys) then storage.get()
for EACH key (N individual HTTP calls re-fetching values one by one).

With 6 entries + contacts + tags, Registratura page fired ~40 sequential
HTTP requests on load, where 3 would suffice.

Fix: Replace list()+N*get() with single exportAll() call in ALL hooks:
- registratura/registry-service.ts (added exportAll to RegistryStorage interface)
- address-book/use-contacts.ts
- it-inventory/use-inventory.ts
- password-vault/use-vault.ts
- word-templates/use-templates.ts
- prompt-generator/use-prompt-generator.ts
- hot-desk/use-reservations.ts
- email-signature/use-saved-signatures.ts
- digital-signatures/use-signatures.ts
- ai-chat/use-chat.ts
- core/tagging/tag-service.ts (uses storage.export())

Additional fixes:
- registratura/use-registry.ts: addEntry uses optimistic local state update
  instead of double-refresh; closeEntry batches saves with Promise.all +
  single refresh
- server-rack.tsx: reversed slot rendering so U1 is at bottom (standard
  rack numbering, per user's physical rack)

Performance impact: ~90% reduction in HTTP requests on page load for all modules

src/modules/hot-desk/hooks/use-reservations.ts

@@ -14,12 +14,11 @@ export function useReservations() {
 
   const refresh = useCallback(async () => {
     setLoading(true);
-    const keys = await storage.list();
+    const all = await storage.exportAll();
     const results: DeskReservation[] = [];
-    for (const key of keys) {
-      if (key.startsWith(PREFIX)) {
-        const item = await storage.get<DeskReservation>(key);
-        if (item) results.push(item);
+    for (const [key, value] of Object.entries(all)) {
+      if (key.startsWith(PREFIX) && value) {
+        results.push(value as DeskReservation);
       }
     }
     results.sort((a, b) => a.date.localeCompare(b.date));