Scaling up read concurrency.

Kafka allows multiple consumers to read from log segments and automatically balances the segments across multiple readers. I think this is a higher level abstraction and introduces all sorts of distributed systems problems if a reader dies and the reading group needs to be rebalanced.

But it would be nice to have some kind of system that could sit on top of these logs and handle that kind of thing. Perhaps using etcd, or some kind of [raft implementation](https://github.com/unshiftio/liferaft).