Update overlay survey script with lessons learned during testnet run

[bboston7]

Description

This change makes a few tweaks to the overlay survey script to fix some small things I noticed after running it on testnet:

• Changes the script's end condition to depend only on responses, and not requests. Without this it was possible for the survey script to run for the full duration of the collecting phase (2 hours) if a node with more than 25 peers stopped responding after the surveyor received the first set of peers.
• Downgrades the severity of "node already in backlog" messages from error to debug. This is an expected condition that I simply forgot to special-case before.
• Modifies the simulator to occasionally return "node already in backlog" messages to test the script against that case.
• Adds a --fast option to the simulate mode that skips any sleep calls. This makes the script much nicer to test.
• Fixes naming of graphml fields to match JSON result fields.
  • I did most of this in the V2 script update, but missed a couple spots.
  • Most of this change is in the simulator to support the new field names.

Checklist

• Reviewed the contributing document
• Rebased on top of master (no merge commits)
• Ran clang-format v8.0.0 (via make format or the Visual Studio extension)
• Compiles
• Ran all tests
• If change impacts performance, include supporting evidence per the performance document