[Diem Data Client] Implement intelligent data summary polling.

[JoshLind]

Today, the Diem Data Client is very primitive: in order to identify the availability of data around it, it samples the storage summaries of peers in a random fashion. As a result, nodes often don't have a sufficient view of the state around them until enough samples have been taken (which may take a non-negligible amount of time).

To address this, we should make the Diem Data Client a little more intelligent:

1. First, when a new peer connects to a node, the Diem Data Client should immediately poll that peer (i.e., to identify the peer's data summary as soon as possible).
2. Second, when regularly polling peers (i.e., to stay up-to-date with changes in data availability), the Diem Data Client should avoid using random sampling and instead provide some form of fairness or guarantees (e.g., a round-robin scheme, or selecting the next peer by identifying the one with the oldest storage summary).

There may also be other optimizations/improvements we can make, but I'll leave that up to the reader 😄.

[JoshLind]

'cc @khiemngo -- something you may be interested in tackling when you get the time! 😄