Make sure that unregistered event sources do not stop JabRef from shu…

[lenhard]

…tting down

Fixes #2486. Unregistering event sources that have not been registered before should not stop JabRef from shutting down orderly. This PR logs such a problem instead of halting the shutdown of the program

• Change in CHANGELOG.md described
• Manually tested changed features in running JabRef

[Siedlerchr]

Would it be possible to first check if the event is registered and only doing a shutdown then?
This should be implemented at best in unregister itself

[lenhard]

@Siedlerchr: Unfortunately, the implementation of the event bus (from Google) supports no such checking functionality. Hence, we would have to remember subscribers in a custom list, which in my point of view would be an unnecessary overhead. However, I have now moved the exception handling code to all the different event sources in JabRef (I hope I found them all). That way, we catch the Exception as early as possible. I cannot think of a circumstance where trying to remove a listener that never listened should crash the complete application.

It is not always logged, though, since not all of the classes that are event sources use a logger. But again, I do not think that this is very tragic, or indeed important.

[Siedlerchr]

There is already discussion about checking it ... google/guava#784
For fixing check style you should add the comments for switching it off at hat position;
http://stackoverflow.com/questions/4023185/how-to-disable-a-particular-checkstyle-rule-for-a-particular-line-of-code?noredirect=1&lq=1

[lenhard]

@Siedlerchr Sorry, but I am not getting your comment regarding checkstyle.

The builds seem to fail because of the usual random fetcher errors.

[Siedlerchr]

@lenhard I meant the codacy, which will complain about the empty catch blocks.
But I think I confused it with checkstyle module

[lenhard]

@Siedlerchr Ok, I understand. However, I am not too fond of codacy and refuse to add comments, just because the tool produces false positives. I am well aware that it is near impossible for a tool to tell if an empty catch block makes sense or not, but it is false positive nonetheless. The existing comment in there is meaningful to humans, which is way more important.

I just pushed a few minor formatting fixes.

[lenhard]

I'll take the "thumps-up" as an "ok-for-merge", which I'll do right away.