rework handling of error, ref #256, #189

[sibson]

revert #189, connectionLost is called after the deferred was called for a successful or failed connection. Leading to AlreadyCalledError being raised

Further investigation uncovered a few bugs

• api.ahutdown() if called multiple times would try to join a dead thread
• renamed functions to be more readable to aid debugging
• ensure we shutdown the reactor and associated threads when exiting program due to an exception. This prevents hangs because the reactor creates a ThreadPool with non-daemon threads
• BREAKING, may now raise Twisted Exceptions rather than VNCDoException wrapping the underlying exceptions

[sibson]

I wonder if using atexit() would be cleaner than what I've done. I'll play with it to see if it works.