Tidy up fault injection scenarios in CI via pumba

[noonio]

• [ ] Write up a few scenarios and check a few scenarios for packet loss
  • [ ] 2 peers; 0%, 50%, 90%, 95% (on one peer)
  • [ ] 2 peers: same again on both peers
  • [x] 3 peers: 0%, 50%, 90%, 9% (one peer)
  • [ ] 3 peers: same again on two peers?
• [ ] Node crashes: TODO: Some scenarios; some amount of downtime
• [ ] Packet delays: ?!
• [x] Be certain pumba/netem is applying (TODO: Try getting tc onto the images directly)
• [x] Be confident that the correct errors are captured by the haskell code/tests
• [ ] ?Stop doing persistence?
• [ ] Update networking page with findings

[noonio]

Closed this ticket in favour of #1592