Regression on Queue tests

[jmesnil]

After r4296[1], we had a lot of regression on the Queue tests[2].

The failing tests were due to a NPE in SimpleString.sizeofString(SimpleString) when encoding a Message with a null Destination.
I fixed the tests by checking if the string was null before computing its size but I'm not sure if it is right in the first place to get a Message with a null Destination.
If that's the case, MessageImpl.encode() should also be guarded against null destination when putting it in the buffer.

I still have 2 failing tests in QueueTest: testChangeConsumersAndDeliver & QueueTest.testMaxSize[3]

[1] http://fisheye.jboss.org/changelog/Messaging/?cs=4296
[2] http://hudson.qa.jboss.com/hudson/view/JBM%202/job/JBM2-all-tests/367/
[3] http://hudson.qa.jboss.com/hudson/view/JBM%202/job/JBM2-all-tests/370/

[timfox]

Destination should never be null.

I'll sort it out today once I've done the ping stuff.