This diff adds an ordered view to the peer set so that highest scores are placed first. This will be useful for selecting the peers we want to get nodes for.
Another diff will cache the score so it's not recomputed every time when it's not needed.
Ref T1696864.