The timer was only started in _logData(), which was called after receiving
the first chunk of data (and only if that was successful), which means
if the first request failed or timed out, the UI would never switch to
a timed out state.

On PX4, under certain circumstances and in rare cases, there can be log
files with 0 size. In that case QGC ignored the entry and tried to
re-request the entry indefinitely.

_finalize methods need information which is not available if all you
have is a mavlink_message_t

Also report the current transfer rate in the currently downloading status to help expose regressions in performance.

I removed some needed headers from a top level file which then causes a
ripple effect downstream