Preserve PeerInfos discovered

[daviddias]

As brought up here: https://github.com/libp2p/js-libp2p/pull/64/files#r108192190

With the DHT, we will want to preserve PeerInfos discovered. This is also the start of having a way to add info about each Peer we know for statistics.

[dignifiedquire]

I would suggest using the peerBook to keep a record of all peers and details we have seen. In addition keeping a simple map peers or connectedPeers which are only {[PeerIdString]: Multiaddr} and record the information about current connections

[daviddias]

Another feature to add is the ability to dump to disk the PeerBook, so that we can shutdown and boot the node and get back almost every connection

[dignifiedquire]

yes, that should be easy to do now that we have the datastore

[dignifiedquire]

we could actually move the peerbook to be a direct wrapper round reading and writing to the datastore instead of in memory that way we consume less memory and it's directly backed up

[daviddias]

@dignifiedquire that would be quite nice! As long as there is some queueing for the writes, so that we don't stress with a 'good to have but not strictly necessary' operation that does lot of IO

[dignifiedquire]

I can easily write datastore wrapper that batches ops if many come in at once :) but those read and writes to leveldb/indexedb are pretty cheap as well.

[dignifiedquire]

Though warning, it will make the whole peerbook api async

[daviddias]

@dignifiedquire we can keep it sync, with an interval that flushes it.

[dignifiedquire]

that doesn't change reads

[daviddias]

@dignifiedquire on repo open, it should just load all the peerBook into memory.

[daviddias]

This has been done \o/