Add in Backoff Filter When Grafting New Peers

[nisdas]

This is a follow up from #473.

When we rejoin a topic in a short duration, behavioural penalties can be triggered if we try to re-graft previously pruned peers. This is prevalent especially if we rejoin a topic before the prune backoff period ends.

• When joining a topic, we make sure we do not graft peers who have a backoff assigned to them. This brings the method in line with the behaviour and peer filter that is applied in our heartbeat method.
• Adds in a regression test for this particular case.

cc @vyzo

[vyzo]

thank you!