Fix actions not being relayed in some circumstances

EnqueuedTelemetryData has actions stored directly in an "actions" field,
but also dependencies, configurations and integrations in separate
fields; extract_telemetry_actions() converts from the latter into proper
TelemetryActions.

However, this method was being called only from a branch of
register_service_and_flush_queued_actions() that would be hit when
RuntimeInfo::apps does not already contain the key corresponding to the
(service, env) in question (so, essentially, deps/confs/integrations
would be lost on all but the first call).

sidecar/src/service/sidecar_server.rs

@@ -644,6 +644,13 @@ impl SidecarInterface for SidecarServer {
                     }
 
                     app.telemetry.send_msgs(actions).await.ok();
+
+                    let mut extracted_actions: Vec<TelemetryActions> = vec![];
+                    enqueued_data
+                        .extract_telemetry_actions(&mut extracted_actions)
+                        .await;
+                    app.telemetry.send_msgs(extracted_actions).await.ok();
+
                     // Ok, we dequeued all messages, now new enqueue_actions calls can handle it
                     completer.complete((service_name, env_name)).await;
                 }