Finish external HTTP client resilience: add shared aiohttp config, retry support, and update task status

backend/app/startup.py

@@ -33,6 +33,22 @@ async def _ensure_database_schema(database_path: str) -> None:
         await db.close()
 
 
+def _create_http_session(settings: Settings) -> aiohttp.ClientSession:
+    """Build a shared aiohttp session with reasonable global limits and timeouts."""
+    timeout = aiohttp.ClientTimeout(
+        total=settings.http_request_timeout_seconds,
+        connect=settings.http_connect_timeout_seconds,
+        sock_read=settings.http_request_timeout_seconds,
+    )
+    connector = aiohttp.TCPConnector(
+        limit=settings.http_max_connections,
+        limit_per_host=settings.http_max_connections,
+        keepalive_timeout=settings.http_keepalive_timeout_seconds,
+        enable_cleanup_closed=True,
+    )
+    return aiohttp.ClientSession(timeout=timeout, connector=connector)
+
+
 async def startup_shared_resources(
     app: FastAPI,
     settings: Settings,
@@ -82,7 +98,7 @@ async def startup_shared_resources(
     if unresolved_count > 0:
         log.warning("geo_cache_unresolved_ips", unresolved=unresolved_count)
 
-    http_session: aiohttp.ClientSession = aiohttp.ClientSession()
+    http_session: aiohttp.ClientSession = _create_http_session(settings)
     geo_service.init_geoip(settings.geoip_db_path)
 
     scheduler: AsyncIOScheduler = AsyncIOScheduler(timezone="UTC")