Contributes to #10

While running e2e tests locally, I detected a hard-to-diagnose bug with vote extensions. This PR contains the fix.

Since I did this in isolation (due to remote work), I'll make an effort at explaining the troubleshooting process in this PR. Hence, this PR will be a succession of commits which show the sketch on how I diagnosed and fixed the bug. So, if the reviewer wants to understand what the problem was and how it was solved, the presence of the little ❌ or ✅ next to each commit below is important (you can click and check how the e2e tests failed/passed).

Additionally, I also found that the e2e tests where blocking forever when a node was crashing. This is fixed in the first commit.

• Commit#1: fix the infinite loop in e2e runner that may block the execution indefinitely.
• Commit#2: adapt ci.toml to only have one validator, since the bug appears when a recovered node has to propose.
• Commit#3: under the hypothesis that something is wrong with VoteExtensionsEnableHeight, we plant a panic to have the backtrace if it happens.
  • The enable height in ci.toml is set to 1007 from genesis, so we should never return zero in this test case.
• Commit#4: contains the fix. Every time we call reconstructLastCommit method of State, immediately after we call updateToState. Therefore the state object whose method we are calling is incomplete. The fix just queries the complete object, which is passed as a normal parameter (and which will be given to updateToState immediately after).
  • Note that, although some checks are failing on that commit (hence the ❌), e2e are now passing
• Commit#5: remove diagnosis-related code, i.e., commit#2 and commit#3

PR checklist

• Tests written/updated
• Changelog entry added in .changelog (we use unclog to manage our changelog)
• Updated relevant documentation (docs/ or spec/) and code comments