Add canonical peer status logging with sampling

[MarcoPolo]

These logs would normally be noisy, so we sample them. And randomly log 1 in N (N is configurable).

This means that most logs are pretty random, but patterns will emerge when there are a large number of similar events.

These logs can then be hooked up to fail2ban to automatically block abnormal peers for a bit.

Example:
A node starts a connection, sends a ping, and closes it. 100 times a second.
We log a connection established at a rate of 1 in 100 (1%).
This would log roughly once a second.
a fail2ban filter can ban a peer after 30 of these logs in a minute. (average of 50 conns/peer/second). And ban that peer for 60 minutes (not sure yet on the timing, but likely anything >10 min is good).

This solution can be applied to similar problems too. Basically anything that boils down to "this is fine/normal if it happens sparingly, but not great if it happens a ton".

Another benefit to this compared to similar solutions is that it's stateless in libp2p. Libp2p doesn't have to do anything besides maybe log a line.