Make the random beacon suck less

[Yawning]

• Implement PVSS
• Rewrite all the consensus bits
  • Switch to unified epochtime/beacon
  • epochtime
  • beacon
• Write the new PVSS node worker
• Alter the tooling to support the new parameters
  • oasis-node genesis
  • oasis-test-runner
• Test
  • Make old tests pass (Blocked on Runtime upgrade E2E test should wait for old node expiration #3404)
  • Add new tests for the trivial failure cases
• Backfill protocol error handling
  • Slashing
  • Worker should retry submitting failed tx(es)
  • The beacon tx(es) should cost gas
  • Disable new node eligibility for next election on end of commit phase
  • Round failure should disable runtimes
• Review/pending changes
  • Unify Threshold/PVSSThreshold
  • Refactoring required to support the transition

To be done later (file issues after merge):

• Make getting the historical epoch suck less (beacon: Make obtaining historical epoch transition points less painful #3667)
• Add more test cases (beacon: Add more test cases #3665)
• Improve performance by adding in the optimizations from the SCRAPE paper. (beacon: Misc. future improvements #3668)
• Add a beacon gRPC service, just so people can query the consensus parameters. (beacon: Add a gRPC service #3666)

I'm not the most thrilled by the performance from the PVSS scheme when used without pairing based crypto, but it has several nice properties:

• It doesn't chew up ridiculous amounts of consensus state, even with a relatively large number of nodes.
• All protocol messages can be independently verified for correctness, on-chain.
• The final entropy generation can be pushed on chain.
• The number of broadcast messages (transactions) is not massive.
• Someone else implemented dleq based PVSS so it's easy to implement.

[codecov]

Codecov Report

Merging #3180 (fc3cc5f) into master (5b261e5) will decrease coverage by 0.58%.
The diff coverage is 59.19%.

@@            Coverage Diff             @@
##           master    #3180      +/-   ##
==========================================
- Coverage   67.38%   66.80%   -0.59%     
==========================================
  Files         390      391       +1     
  Lines       36687    38122    +1435     
==========================================
+ Hits        24723    25468     +745     
- Misses       8511     9000     +489     
- Partials     3453     3654     +201     

Impacted Files	Coverage Δ
go/common/node/node.go	65.48% <ø> (ø)
go/consensus/api/api.go	14.28% <ø> (ø)
go/consensus/tendermint/abci/mux_mock.go	0.00% <ø> (ø)
go/consensus/tendermint/api/state.go	60.81% <ø> (ø)
...onsensus/tendermint/apps/governance/state/state.go	73.23% <ø> (ø)
...consensus/tendermint/apps/keymanager/keymanager.go	66.25% <ø> (ø)
go/consensus/tendermint/apps/staking/query.go	76.92% <ø> (+3.84%)	⬆️
...nsensus/tendermint/apps/staking/signing_rewards.go	20.00% <ø> (ø)
go/consensus/tendermint/apps/staking/staking.go	51.70% <ø> (-1.37%)	⬇️
...nsus/tendermint/apps/supplementarysanity/checks.go	42.78% <0.00%> (+1.06%)	⬆️
... and 113 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5b261e5...fc3cc5f. Read the comment docs.

[kostko]

Should we make this required in master? And optional in the backported version.

[Yawning]

Depends on how the transition will happen. We could save a bunch of backporting effort if the transition is just "Deploy the new code with the old backend enabled, then switch backends at a later date". If that is the option taken, then this needs to remain optional.