feat: wire up storage latency metrics, uptime gauge, and config timeouts

Instrument DeliveryService (enqueue, fetch) and KeyService
(key_package_upload, key_package_fetch) with storage latency histogram
recording. Add periodic uptime gauge task (every 15s). Log effective
rpc_timeout_secs, storage_timeout_secs, and webtransport_listen at
startup to eliminate dead_code warnings on EffectiveConfig fields.

crates/quicprochat-server/src/domain/delivery.rs

@@ -80,12 +80,14 @@ impl DeliveryService {
         let mut first_seq = 0;
 
         for (i, dk) in device_keys.iter().enumerate() {
+            let start = std::time::Instant::now();
             let seq = self.store.enqueue(
                 dk,
                 &req.channel_id,
                 req.payload.clone(),
                 ttl,
             )?;
+            crate::metrics::record_storage_latency("enqueue", start.elapsed());
             if i == 0 {
                 first_seq = seq;
             }
@@ -106,12 +108,14 @@ impl DeliveryService {
     /// The `recipient_key` should be the device-scoped composite key
     /// (`identity_key + device_id`) or bare `identity_key` for single-device.
     pub fn fetch(&self, req: FetchReq) -> Result<FetchResp, crate::storage::StorageError> {
+        let start = std::time::Instant::now();
         let messages = if req.limit > 0 {
             self.store
                 .fetch_limited(&req.recipient_key, &req.channel_id, req.limit as usize)?
         } else {
             self.store.fetch(&req.recipient_key, &req.channel_id)?
         };
+        crate::metrics::record_storage_latency("fetch", start.elapsed());
 
         Ok(FetchResp {
             payloads: messages