Schedule re-subscription if necessary after node restart

[acogoluegnes]

Necessary = queue does not exist if classic queue, durable,
and in a cluster setup. In such a case, the queue "does
not exist" as long as the node is down and becomes available
again when the node restarts. Consumers on other nodes
are then scheduled to subscribe for some time.

Fixes #330

[acogoluegnes]

Acceptance

Start a cluster from the umbrella:

make start-cluster PLUGINS="rabbitmq_management"

Get PerfTest and start it:

cd /tmp
git clone git@github.com:rabbitmq/rabbitmq-perf-test.git
cd rabbitmq-perf-test
git checkout rabbitmq-perf-test-330-some-classic-queues-do-not-have-consumers-after-node-restart
make binary
tar xf target/rabbitmq-perf-test-2.18.0-SNAPSHOT-bin.tar.gz
rabbitmq-perf-test-2.18.0-SNAPSHOT/bin/runjava com.rabbitmq.perf.PerfTest -x 10 -y 10 -r 10 -c 10 \
    -q 10 -qp 'cq-%d' -qpf 0 -qpt 9 --auto-delete false -f persistent \
   -H amqp://guest:guest@localhost:5672,amqp://guest:guest@localhost:5673,amqp://guest:guest@localhost:5674

Go to http://localhost:15672/#/queues, it should look like the following (10 queues, 10 msg / second each, 1 consumer each):

From the umbrella directory, stop and restart the first node:

./sbin/rabbitmqctl -n rabbit-1 stop_app; ./sbin/rabbitmqctl -n rabbit-1 start_app

After a few seconds, PerfTest should resume the consumers that went down: the consumed rate should still be 100 msg / second overall and the queues page should show that each queue has 1 consumer.