* maintain pings in a hashmap
* convert `KeepAliveManager` into a singleton
* chore: fix an unrelated cyclic dependency error
* update `selectPeerForProtocol` to return peer with the lowest latency
* use the new KeepAliveManager API
* use the new API for `selectPeerForProtocol`
* add tests
* use PeerData to hold the ping instead of a new variable
* improve tests for readability
* move back KeepAliveManager from singleton
* reenable all tests
* minor improvements
* improve error handling
* convert .then() syntax to async/await