Merge datastore and libp2p-peerstore

[tomaka]

There's no reason to have datastore in its own crate.

[twittner]

While at it, you would not be interested in twittner:datastore-queries, wouldn't you?

[adamski]

I'm pretty much lurking here at the moment, but would if these were merged would it affect the ability to plug in different underlying data stores?

[tomaka]

Right now there's a JSON datastore, a memory datastore, a JSON peerstore, and a memory peerstore.
Removing the datastores doesn't remove the possibility to choose the way you want to store your peers.

The problem of the datastore crate is that its design is way too generic, and therefore inefficient. The memory datastore is nothing more than a hashmap, and the JSON datastore simply loads everything into a hashmap because there's no way it can do an efficient lookup in a file when we need a specific entry.

[adamski]

Yes, what I'm looking for eventually is being able to have a datastore that can be efficiently queried locally, basically a local KV/JSON database.

[tomaka]

No longer relevant after #723