Recovery related changes

[vikinghawk]

This PR contains some changes related to recovery. Note i cleaned up a bunch of compiler warnings by adding the missing Override tags on methods. I can submit another PR without those changes if you guys would prefer.

Non-cosmetic changes are in AutorecoveringConnection and AutorecoveringChannel.

Fixes:

1. fixes a NPE that can occur when a thread attempts to create a new channel after a connection has been recovered but before channel recovery has completed. More details of issue at https://groups.google.com/forum/#!topic/rabbitmq-users/7P5fYQLLP6g. Fixes the NPE in AutorecoveringChannel, but then also changed the AutorecoveryingConnection to not set the connection delegate until channel recovery is complete. This allows channel recovery to complete without error at the expense of the separate thread attempting to create a new channel of getting an AlreadyClosedException because it would happen on the old delegate.

Other changes:

1. Added a check in recoverQueues() that skips over all the queue name change logic if the name remained the same.

[michaelklishin]

Thank you!

[michaelklishin]

This method existed for a reason. Are you 100% sure it is safe to remove it now, or a good idea to do so?

[michaelklishin]

Digging in code history suggests that the method was introduced around the time we wanted to move Langohr and March Hare to use recovery implemented by this client, and that's why 1e603d9 calls out Langohr specifically.

So far it looks like it's safe to delete this method.

[michaelklishin]

OK, Langohr recovery test suite works just fine with this change.

[acogoluegnes]

@vikinghawk Thanks!

[michaelklishin]

This was back ported for 3.6.6 => changing milestone.