Optimise NodeConnectionManager.getClosestGlobalNodes to concurrently contact nodes

[tegefaulkes]

Specification

getClosestGlobalNodes is the core logic for finding a node within the Kademlia network. Currently is implemented as a priority queue that queries a node for a list of nodes close to the target node. it then sequetially pings each node to see if it's alive before adding it to the NodeGraph.

This needs to be optimised to ping K nodes concurrently. It also needs to only ping and contact nodes in the priority order.

Additional context

• Related Integrate TaskManager into NodeGraph and Discovery #445
• Related Integrate TaskManager into NodeGraph and Discovery #445 (comment)

Tasks

• 1. Nodes need to be pinged concurrently to check if they're alive. We can use the Semaphore from @matrixai/async-locks to enforce the concurrency limit.
• 2. Only ping a node just before trying to connect to it. There is no need to ping every node since we only need to connect in the priority order.

[tegefaulkes]

This issue is not relevant anymore. The findNode procedure was completely revamped in #618 .

[CMCDragonkai]

So it is not sequential anymore?

[tegefaulkes]

Yep, up to 3-6* connections are made concurrently during the nodesFind operation depending on how it's configured. Nodes are added to the NodeGraph when the NodeManager handles the connection event.