Do not call Debugf for non-debug levels in doDropRPC

[algorandskiy]

In high load scenarios when consumer is slow, doDropRPC is called often and makes extra unnecessary allocations invoking log.Debug.

Fixed by checking log level before running expensive formatting.

Before:

BenchmarkAllocDoDropRPC-10    	13684732	        76.28 ns/op	     144 B/op	       3 allocs/op

After:

BenchmarkAllocDoDropRPC-10    	28140273	        42.88 ns/op	     112 B/op	       1 allocs/op

[vyzo]

nice one.

[vyzo]

Can you go mod tidy please?

[algorandskiy]

done.

There is also extra allocation in Message.GetFrom() on []byte to peer.ID conversion (runtime.slicebytetostring) but I'm not sure how to fix it properly. One approach could be caching []byte -> peer.ID on a first call and reuse on subsequent calls. Another one is to use []byte instead of peer.ID everywhere but it does not look right (although I'm not sure why peer.ID is a string and not []byte as it contains binary data).

[vyzo]

peer.ID is a string so that you can easily compare and use as a key in maps, there is a lot of convenience to that.

[algorandskiy]

I updated the PR title/description to properly denote the source of extra allocations: it is strings escaping b/c of interface{} in Debugf signature, and not the message formatting itself that happens later inside Debugf after checking the log level (I've confused it with fmt.Sprintf() in doDropRPC callers.