Separate election from ticket generation (Part of #2225)

[ZenGround0]

Why do we need this PR

Right now ticket generation and election proof generation are one and the same. To get closer to the planned consensus construction we need to separate those things out. This PR does so while at the same time tying the ticket and election proof operations to objects to organize functionality together.

• TicketMachine and ElectionMachine refactor
• Refactor consensus.Expected and mining.Worker to use these dependencies instead of consensus free functions
• Fix mining.Worker to use separate ticket generation and election proving operations

Non-goals

• Introduce look back parameter for election proofs to finish out EC Upgrade 4: integrate randomness sampling into the leader election rules #2225 (this would require fake work if done before EC Upgrade 2: randomness sampling from ticket chain #2223)
• Semantic ticket generation correctness, i.e. multiple tickets per block when mining null blocks. As a stop gap this PR includes a null block count parameter in ticket validation and generation.

[ZenGround0]

@anorth @frrist I've put in effort to make this PR's commit history helpful for review. It is best reviewed one commit at a time.

• The first commit introduces the fundamental refactor that puts election proof and ticket generation and validation on separate objects
• The second commit is a prefactor that turned out to be necessary in order to maintain existing structure. It imperfectly adds a call to determine the miner worker address to the power table. While this is not very clean it only messes with an abstraction that is arguably not that clean already (the power table).
• The third commit integrates the new election and ticket validation logic into the consensus validation rules checking
• The fourth commit integrates this logic into the mining worker. It removes the dependency-injection doSomeWorkFunc because now these dependencies can be injected using the ticket generation interface. It also adds a clock as a worker dependency so that sleep calls here can be made using a system aware clock.
• The fifth commit integrates all of the above into the node. A lot of work went into getting node tests to pass. Mocking out fake tickets and election proofs was challenging enough that I opted to actually mine winning blocks in process to use for the tests. We'll need something better soon.

[anorth]

This is one of those calls that should be clearly parameterised by the tipset keying the state in which the address is looked up, #3022. Can you make that change in a prefactor?

[ZenGround0]

#3401 addresses this. That may merge first anyway, but I think this PR can go in before that if this one is ready to go first as the race #3401 prevents is rare and is unlikely to cause test flakiness / bad user behavior in a short amount of time.

[ZenGround0]

Merging #3401 is proving tricky. Deferring to an immediate follow on in favor of getting this in for 0.5 staging.

[anorth]

My feedback is mostly minor: great work. The opaque hex strings are perhaps my biggest problem.

The factoring here definitely needs some work so that we can write much more thorough tests, but well done on improving the situation.

[codecov-io]

Codecov Report

Merging #3287 into master will increase coverage by <1%.
The diff coverage is 65%.

@@           Coverage Diff           @@
##           master   #3287    +/-   ##
=======================================
+ Coverage      47%     47%   +<1%     
=======================================
  Files         229     230     +1     
  Lines       14254   14305    +51     
=======================================
+ Hits         6751    6850    +99     
+ Misses       6513    6444    -69     
- Partials      990    1011    +21

[frrist]

This looks great @ZenGround0!