Allow setting bucket size to 16 or higher

[asdacap]

• It does not look like our discv4 implementation implement a complete walk like geth, meaning we are limited by the nodes that are available from the nodes that are in our buckets, which does not get replaced (it can happen, but its very rare that they don't reply a pong). And because of how the distance calculation works, only bucket 245ish and above is used. So it usually keep getting largely the same set of neighbour. Eventually it'll get enough peer to fill the 10000 default candidate list, but it take some time.
• The only way to increase the range of node that it discover without a significant rewrite is to increase the bucket size, which is what this PR do, and also expose the option to increase further it in case of smaller network.

Changes

• Set back discovery bucket size to 16, which is what the spec said.
• Set limit of neighbour msg to 12, which prevent it from becoming too big in case of ipv6 address.
• Allow setting discovery config from CLI.

Types of changes

What types of changes does your code introduce?

• Optimization

Testing

Requires testing

• Yes
• No

If yes, did you write tests?

• Yes
• No

Documentation

Requires documentation update

• Yes
• No

Requires explanation in Release Notes

• Maybe

Improved peer discovery

[LukaszRozmej]

Maybe a significant rewrite would be desired? This code never got the love.

[asdacap]

Maybe, but @flcl42 is already working on discv5, so this is getting deprecated anyway. WDYT?

[LukaszRozmej]

Maybe, but @flcl42 is already working on discv5, so this is getting deprecated anyway. WDYT?

Not sure if it will

[LukaszRozmej]

Is there any negative effect on memory or performance?

[asdacap]

Hmm, I guess it would have more entry in the bucket. But assuming only bucket 245-255, that would only be 160 record. Its very low. Too low even.

[LukaszRozmej]

Hmm, I guess it would have more entry in the bucket. But assuming only bucket 245-255, that would only be 160 record. Its very low. Too low even.

Can you check some profiling?

[asdacap]

Run is 16, 12, 16, 12, 16, 12. Not correlation. Nethermind hooked to shutdown on 8000 peer in peer pool. There are other modification to make it get 8000 peer in less than 4 minute. It normally take 30 minute maybe.