swarm/network: WIP consider all nodes for healthy iteration

[nolash]

addresses ethersphere/swarm#1263

This PR corrects two problems with the existing WaitTillHealthy implementation.

1. The healthy check only called ConnectNN, should instead call Healthy()
2. The kademlias being checked were only from the nodes that were up at the time. It should instead include all nodes that are in the simulation.

[acud]

LGTM except for a few code-style comments.
The strange thing about this is that I had a feeling that the AddNodes and ConnectNodesX would actually block. Judging by the results of the snapshot it means that both these methods return and potentially before the nodes were even up? This is quite strange. Moreover considering that the Node.Up is set to be true when creating the node - they should all be up after create.go#L82. We should add a panic in Net.ConnectX when the Up node count does not equal the parameter value.

[acud]

can be removed

[holisticode]

why?

[zelig]

@holisticode there is a consistent CI failure on snapshot creation
https://travis-ci.org/ethereum/go-ethereum/jobs/497949535#L816

[zelig]

should just use AddNodesAndConnectRing here