Offer explicit API on SWIM.Instance for "start monitoring this node"

[ktoso]

Expected behavior

Shells today kind of work around this and implement it manually, each in its own slightly different way.

E.g. SWIMNIOShell does:

        let sequenceNumber = self.swim.nextSequenceNumber()
        targetPeer.ping(payload: self.swim.makeGossipPayload(to: nil), from: self.peer, timeout: .seconds(1), sequenceNumber: sequenceNumber) { (result: Result<SWIM.PingResponse, Error>) in

which is a bit hacky, since we have to reach into the instance more than we really should, both making a sequence number and a gossip explicitly should not be needed.

Alternatively, this can be solved via swim.addMember and just blindly adding it there, so swim starts monitoring it.

This honestly is probably just fine, but I was wondering if we should just expose the addMember or not, as in some impls where nodes do not have known UID before this adds more complexity than kicking off a ping which will return with the UID.

Actual behavior

Some explicit call to be made on the isntance which tells us how to deal with "start monitoring that node", including retries perhaps etc?

Version/commit hash

0.1.0