fix race condition in grpc streams

Write should not be called again before it is finished.
That can happen with our thread pool, hence the mutex.

It seems like it was not crashing before, and the recent
update of grpc made it visible when running on iOS.

src/backend/src/core/core_service_impl.h

@@ -21,14 +21,20 @@ class CoreServiceImpl final : public mavsdk::rpc::core::CoreService::Service {
         const rpc::core::SubscribeConnectionStateRequest * /* request */,
         grpc::ServerWriter<rpc::core::ConnectionStateResponse> *writer) override
     {
-        _dc.register_on_discover([&writer](const uint64_t uuid) {
+        std::mutex connection_state_mutex{};
+
+        _dc.register_on_discover([&writer, &connection_state_mutex](const uint64_t uuid) {
             const auto rpc_connection_state_response = createRpcConnectionStateResponse(uuid, true);
+
+            std::lock_guard<std::mutex> lock(connection_state_mutex);
             writer->Write(rpc_connection_state_response);
         });
 
-        _dc.register_on_timeout([&writer](const uint64_t uuid) {
+        _dc.register_on_timeout([&writer, &connection_state_mutex](const uint64_t uuid) {
             const auto rpc_connection_state_response =
                 createRpcConnectionStateResponse(uuid, false);
+
+            std::lock_guard<std::mutex> lock(connection_state_mutex);
             writer->Write(rpc_connection_state_response);
         });