Improvements in Requests article

[slowli]

The Requests article was merged into master in #35, but some questions were left unanswered:

• How do timeouts in requests influence the theoretical properties of the consensus algorithm, given the assumption about the partial synchronicity of the network? (Kudos to @YuryYa for raising the question.)

  • Action item: a note in the requests article once this is figured out.

• How does the processing of messages/states described in Requests corresponds to one described in the consensus spec from #36? Are they interleaved, or is one performed after another, or smth else?

  • Action item: description in the spec and the requests article.

@YuryYa, please feel free to add your suggestions as well.