New staking reward system

[kianenigma]

Objectives:

1. Staking roles (ledger, etc.) to be entirely decoupled from reward-related logic.
2. (probably not worth it) Rewards to be mire customizable on a per-nominator basis. e.g. one nominator not being paid out automatically.
3. Rewards to be scalable: they should be payable in smaller chunks. probably requires multi-page exposure table.
4. Once we have 3, it should be very simple to actually make rewards automatic. Even better, automatic for those who want it to be.

Some this might require us to make things seemingly less efficient, but if we do it multi-page, it will actually be more scalable.

[kianenigma]

related: #473 #282

[burdges]

I'll remark that availability and approval reward points flowing into the validator end up kinda nuanced, but they still have a (era,point) resolution in the end. https://hackmd.io/@rgbPIkIdTwSICPuAq67Jbw/rkTzZX_8t

[Ank4n]

I am tackling point 1 and 3 of the issue in the PR paritytech/substrate#13498.

Since the reward would be distributed in multiple pages, one of the objectives of this PR is to incentivise a validator to pay out all pages of the reward. My current approach is to distribute commission of the validator across pages proportional to the total stake of the page.

I had couple of great suggestions for improvement and I wanted to get some opinion from @kianenigma and @burdges as well on this before making changes.

Suggestion 1 (@gpestana):
Instead of commission proportional to page stake, it can be evenly distributed across all pages irrespective of page stake. We can extend this even further and pay the validator stake also evenly across pages. (Right now validator stake is paid out in 1st page).

Suggestion 2 (@rossbulat) :
We can reverse the order of rewards payout, so that the page with lowest stake is paid out first and the highest stake is paid out only at last incentivising validators to pay out all pages. There is still an option though to pay out a page explicitly instead of pallet choosing the order automatically.

I am leaning towards doing the Suggestion 1 which would make Suggestion 2 not needed. What do you guys think?

[kianenigma]

I think all of these suggestions are coming the perspective of protecting small nominators who are in the lesser pages from not receiving their rewards. This is fair concern, but I would generally avoid designing systems too much with such a mindset in mind. The goal of the game theory involved in FRAME or blockchains is not necessarily to protect everyone at any cost, but rather be a system that is sound, sustainable, or foremost unstoppable.

I think if we are to solely make decisions from this perspective, then the existing model makes perfect sense; if the total stake in a page is so small that no one has an incentive to claim it, then it should not be claimed (recall that anyone can do the claims, not just the validator).

That all being said, I won't resist against tweaking this system such that we slightly protect nominators, in which case I like option 1 way better. It doesn't 100% solve the issue, but it will drastically reduce the chances of a page being left unpaid.

[kianenigma]

@Ank4n @gupnik @gpestana this is all done, except the last item. Closing. Please make an issue about the proposal to use #[pallet::task] or on_idle for optional automatic rewards.