Additional logging and friendlier error messages

[tomusdrw]

Probably solves #848 also - when compiled in debug it takes significant amount of time between Closing... message and stopping the application (my cpu goes high during that time also) - I will check backtraces what's happening in that particular moment, but having hard time to reproduce stalling after Ctrl+C.

[arkpar]

The delay is due to importing 128 blocks at a time in client.rs There is no early exit in this loop I believe

[tomusdrw]

@arkpar Could you have a look at this? Does it make any sense to shutdown event_loop before clearing workers?

clear() drops workers and locks until they process all work, but my guess is that some messages might still be added to the channel for them if we don't shutdown event_loop?

[arkpar]

If event loop is destroyed while workers are active they might panic trying to send notification to the event loop. Any particular reason to remove this?
If there are unhandled events left in the event loop that's fine cause we just ignore them. We don't have to process anything after Shutdown event.

[tomusdrw]

Ok, reverting. I thought that because event_loop is not shut down workers may still get additional work because event_loop is active. My understanding was that swapping those two lines will block any new events from event_loop and after that we will wait for workers to process what's pending.