Update networking section

[lamafab]

Fixes #234.

Things to consider:

• The Kademlia/discovery process is quite thin. Based on a conversation with @infinity0, we should just refer to the libp2p website and spec the details of any questions the other implementers come up with.
• Currently, we just refer to external specs regarding libp2p, noise, mplex/substreams. Unless we try to reimplement this from scratch, we probably need to rely on the feedback of implementers for what must be expanded.

[tomaka]

Grandpa and authority-discovery are missing, but I guess it's intentional for now

[drskalman]

Reviewed Appendix E. Comments are in a0c8160

[drskalman]

Appendix e reviewed by @drskalman. After @lamafab goes over the changes, @FlorianFranzen will review the result again.

[FlorianFranzen]

Very nice work on consolidating all these different sources and getting rid of all these outdated chapters and appendices.

I would just like to ask you to make it a bit more clear how the protocol identifiers in the sections "Connection establishment" and "Protocols and Substreams" are related, including defining what "After the protocol is negotiated" means and which substreams uses which type of substream protocols.

It might also be useful to mention that protobuffer encodes each fields as a key value pair using the provided id, etc, but it is hardly a must. People can always read the official documentation.

[lamafab]

@FlorianFranzen The "protocol identifiers" section was removed from the spec, since it's not directly relevant. This spec is exclusively about Polkadot and those identifiers are essentially just prefixes on substreams. I did add a note in the "Protocols and Substreams" section about it, however.

Each substream now mentions the substream type (Request-Response or Notification substream).

[FlorianFranzen]

Great. Thank you for getting these last things fixed. LGTM!

[FlorianFranzen]

Very nice. I am sure that was not fun to rebase. Just a few smaller things left to fix.