Add EntityPreloader for batched profile metadata preloading

Implements an actor-based preloading system to efficiently fetch profile
metadata for note authors and referenced users. The EntityPreloader queues
requests and batches them intelligently (500 pubkeys or 1 second timeout)
to avoid network overload while improving UX by ensuring profiles are
available when rendering notes.

Key changes:
- Add EntityPreloader actor with queue-based batching logic
- Integrate with SubscriptionManager via PreloadStrategy enum
- Add lifecycle management (start/stop on app foreground/background)
- Skip preload for pubkeys already cached in ndb
- Include comprehensive test suite with 11 test cases covering batching,
  deduplication, and edge cases
- Optimize ProfilePicView to load from ndb before first render

Closes: https://github.com/damus-io/damus/issues/gh-3511
Changelog-Added: Profile metadata preloading for improved timeline performance
Signed-off-by: Daniel D’Aquino <daniel@daquino.me>

damusTests/NostrNetworkManagerTests/NostrNetworkManagerTests.swift

@@ -224,7 +224,7 @@ private final class MockSubscriptionManager: NostrNetworkManager.SubscriptionMan
         super.init(pool: pool, ndb: ndb, experimentalLocalRelayModelSupport: false)
     }
 
-    override func streamIndefinitely(filters: [NostrFilter], to desiredRelays: [RelayURL]? = nil, streamMode: NostrNetworkManager.StreamMode? = nil, id: UUID? = nil) -> AsyncStream<NdbNoteLender> {
+    override func streamIndefinitely(filters: [NostrFilter], to desiredRelays: [RelayURL]? = nil, streamMode: NostrNetworkManager.StreamMode? = nil, preloadStrategy: NostrNetworkManager.PreloadStrategy? = nil, id: UUID? = nil) -> AsyncStream<NdbNoteLender> {
         let lenders = queuedLenders
         return AsyncStream { continuation in
             lenders.forEach { continuation.yield($0) }