R4R: F1 fee distribution

[cwgoes]

Closes #2916

Implement the F1 spec in the "simple form" (inflation is left separate for the time being).

This PR is structured to be reviewed as a sequence of commits, commit-by-commit.
Each is independent (builds / passes simulations successfully, except for the import/export test which I left for the end).

1. The first commit removes old fee distribution code which will no longer be used
2. The second commit updates PENDING.md ( 😉 )
3. The third commit implements the basic types, file layout, begin block allocation, and supply tracking.
4. The fourth commit implements all the requisite hooks, period tracking, withdrawals
5. The fifth commit implements a few bugfixes and simulation testing.

Saved for a separate PR: queriers and CLI commands to view state (mostly reincarnated from last time). I can add that as a PR to this one if we want to wait on those to merge, but easier to review just the substance first I think.

In my view, this version of fee distribution compares favorably to our current one: it is not approximate (so e.g. the whole class of concerns illuminated by #2764 are eliminated), it completely separates withdrawal for separate users (no delegator or validator can cause another to be forced to withdraw fees, thus eliminating the class of concerns of #2914), it makes it far easier for us to change the distribution function per-validator (so e.g. #2525 and #3023 can more easily be addressed), it is implementationally simpler, and it is more amenable to future improvements (see the spec).

One concern is potential error introduced by utilizing decimals - although F1 is analytically exact, we cannot use arbitrary-precision rational numbers in the state machine and must make do with decimals.

The validator update delay renders fee payments slightly inaccurate (e.g. delegations will be paid a block before their validator actually signs), but I don't think this impacts the implementation. The most complex / nuanced part is probably dealing with slashes, which must be iterated through on withdrawal.

F1 has not yet been tested on a testnet. Depending on results of review, given our current timeline, this seems like a candidate for "Red Wedding" (#3079).

• Linked to github-issue with discussion and accepted design OR link to spec that describes this work.
• Wrote tests
• Updated relevant documentation (docs/)
• Added entries in PENDING.md with issue #
• rereviewed Files changed in the github PR explorer

---

For Admin Use:

• Added appropriate labels to PR (ex. wip, ready-for-review, docs)
• Reviewers Assigned
• Squashed all commits, uses message "Merge pull request #XYZ: [title]" (coding standards)

[codecov]

Codecov Report

Merging #3099 into develop will increase coverage by 0.16%.
The diff coverage is 18.46%.

@@            Coverage Diff             @@
##           develop   #3099      +/-   ##
==========================================
+ Coverage    55.64%   55.8%   +0.16%     
==========================================
  Files          134     132       -2     
  Lines         9746    9488     -258     
==========================================
- Hits          5423    5295     -128     
+ Misses        3980    3858     -122     
+ Partials       343     335       -8

[alexanderbez]

Did a first run through and left some comments. Overall, looks really good 👍

[bneiluj]

So there will be no need to implement a specific Auto Withdraw + Delegate fee script for validators ?

[cwgoes]

So there will be no need to implement a specific Auto Withdraw + Delegate fee script for validators?

This instantiation of F1 does not include automatic inflation calculation, so re-bonding will still be necessary. Future versions may implement inflation per the spec, which would render it unnecessary.

[bneiluj]

Got it. Thanks.

[rigelrozanski]

Generally a very clean implementation 👏

Main considerations:

• currently the file organization is unintuitively shuffled for me - within keeper, we really ought to reorg functions into files relating to either objects (aka delegation.go, validator.go) or abci calls (tx.go, end_block.go etc. )
• use of sdk.Dec for staking should be removed (see comment)
• lack of NewXxx functions
• there is a missing test which would ease my heart.

[rigelrozanski]

🏁 🏁 🏁 🏎
🏅

[jackzampolin]

Ready for merge! Great work @cwgoes and @cosmos/cosmossdk team!