Implement proper routing with fallback

[dkropachev]

Introduce configuration primitives and implement logic that will allow users to have fallback logic for routing that enables following scenarios:

1. User wants to target particular rack, with fallback to another rack
2. Target rack with fallback to dc
3. Target dc with fallback to whole cluster
4. Target rack with fallback to dc with fallback to cluster

Fallback means two things:

1. Configuration validation logic should honor fallbacks, if no nodes reported by current routing scope, it will get nodes from the fallback scope, repeating this proccess untill fallback is nil
2. When it known updates nodes from alternator it does the same thing

When it switches to fallback it should throw a log message.

Naming:
RoutingScope: rack, datacenter or cluster