Fix new metric where we used ms instead of seconds (#7771)

Introduced in #7755, not yet released.

changelog.d/7771.misc

+Add some metrics for inbound and outbound federation latencies: `synapse_federation_server_pdu_process_time` and `synapse_event_processing_lag_by_event`.

synapse/federation/sender/__init__.py

@@ -209,7 +209,7 @@ class FederationSender(object):
 
                         synapse.metrics.event_processing_lag_by_event.labels(
                             "federation_sender"
-                        ).observe(now - ts)
+                        ).observe((now - ts) / 1000)
 
                 async def handle_room_events(events: Iterable[EventBase]) -> None:
                     with Measure(self.clock, "handle_room_events"):

synapse/handlers/appservice.py

@@ -118,7 +118,7 @@ class ApplicationServicesHandler(object):
                         ts = yield self.store.get_received_ts(event.event_id)
                         synapse.metrics.event_processing_lag_by_event.labels(
                             "appservice_sender"
-                        ).observe(now - ts)
+                        ).observe((now - ts) / 1000)
 
                     @defer.inlineCallbacks
                     def handle_room_events(events):