perf: offload per-NIO ubridge commands to thread-pool executor

Replace the 3-5 sequential await _ubridge_send calls in _connect_nio
with a single run_in_executor batch.  The batch holds the node-level
asyncio Lock to prevent interleaving with async sends, then uses the
hypervisor's new send_batch_sync method which does blocking socket
sendall/recv inside the thread pool.  Different nodes' batches now
run in true OS-thread parallelism rather than serialising through
the asyncio event loop between every command.

- ubridge_hypervisor.send_batch_sync: blocking batch send using
  the underlying socket from the asyncio transport, protected by
  threading.Lock.
- _connect_nio: builds command list (add_nio_udp, start_capture,
  bridge start, reset_packet_filters, add_packet_filter) and
  dispatches to the default executor.

tests/compute/docker/test_docker_vm.py

@@ -1440,9 +1440,12 @@ async def test_add_ubridge_connection(vm):
         call.send('bridge create bridge0'),
         call.send("bridge add_nio_tap bridge0 tap-gns3-e0"),
         call.send('docker move_to_ns tap-gns3-e0 42 eth0'),
-        call.send('bridge add_nio_udp bridge0 4242 127.0.0.1 4343'),
-        call.send('bridge start_capture bridge0 "/tmp/capture.pcap"'),
-        call.send('bridge start bridge0')
+        call.send_batch_sync([
+            'bridge add_nio_udp bridge0 4242 127.0.0.1 4343',
+            'bridge start_capture bridge0 "/tmp/capture.pcap"',
+            'bridge start bridge0',
+            'bridge reset_packet_filters bridge0',
+        ]),
     ]
     assert 'bridge0' in vm._bridges
     # We need to check any_order otherwise mock is confused by asyncio