AP_HAL_SITL: process inbound data in outqueue-length delay loop

this is the loop which ensures the amount of data sent to the mavlink client (usually Python) is limited - if we don't do this then we lose vast amounts of data when running at large speedups.

By attempting to process inbound data we may realise that the TCP connection has been dropped, and in that case we will start to listen for another connection.

This allows you to terminate the sim_vehicle.py MAVProxy and have it automagically restart (when running under GDB).  This is very useful for testing MAVProxy patches with SITL; it's a different workflow to opening an output and connecting a new version of MAVProxy to that outout.

libraries/AP_HAL_SITL/SITL_State.cpp

@@ -185,12 +185,14 @@ void SITL_State::wait_clock(uint64_t wait_time_usec)
     // conditions.
     if (speedup > 1 && hal.scheduler->in_main_thread()) {
         while (true) {
-            const int queue_length = ((HALSITL::UARTDriver*)hal.serial(0))->get_system_outqueue_length();
+            HALSITL::UARTDriver *uart = (HALSITL::UARTDriver*)hal.serial(0);
+            const int queue_length = uart->get_system_outqueue_length();
             // ::fprintf(stderr, "queue_length=%d\n", (signed)queue_length);
             if (queue_length < 1024) {
                 break;
             }
             _serial_0_outqueue_full_count++;
+            uart->handle_reading_from_device_to_readbuffer();
             usleep(1000);
         }
     }