Track reward state

[sandtreader]

Create a new RewardState module which tracks the status of reward accounts. To do this, it needs to:

• Track delegation of stake to SPOs
• Track block production by SPOs
• Calculate total fees for each block
• Calculate reward split based on those fees and SPO data
• Take account of deposits for staking and governance

It should output a list of reward deltas for each block. We might consider making this the same format as address deltas, and reuse the AddressState module to integrate it.

[sandtreader]

Note: It doesn't need to track the delegation of stake to SPOs. The SPOState will be doing that but I don't think the stake delegation is needed anyway, because it governs who is supposed to produce a block - we are only interested who actually did produce a block at this stage.

^WRONG: Links below say that relative performance (the number of blocks produced / the number of blocks that should be produced given stake weight) does matter.

[sandtreader]

This looks like it should be a gold mine - as I understand it they have implemented all the wrinkles at least up until the first GA change

https://github.com/cardano-foundation/cf-java-rewards-calculation

[sandtreader]

Useful reference with a big formula!

https://docs.cardano.org/about-cardano/learn/pledging-rewards

[sandtreader]

Gotcha's flagged by the Amaru team at the Paris workshop:

• The SPD used for block header validation is captured including rewards at epoch 'e' for use in epoch 'e+2'. But the changes to the treasury pots only happen at the start of epoch 'e+2', so there is a two epoch window where the totals don't add up
• There was a bug before Allegra (one of the protocol versions, forget which) where if two SPOs delegated to the same stake, their rewards weren't added together but one overwrote the other, order dependent. It's not clear what the ordering was, but it's in the Haskell code for the previous era, and I think the Java code above does it (they note this in their description). We need to implement both versions, depending on the protocol version (* which we need to capture in the BlockInfo). Sam Leathers said there were some MIRs (explicit movements from treasury) to compensate the SPOs that missed out.
• Missing credentials - if an SPO retires or stake delegation retires between capture at epoch 'e' and use in epoch 'e+2', their rewards go to the treasury

[sandtreader]

Moved to AccountsState (may become StakingState, depending where the treasury pots end up) - this ticket now limited to generating the rewards in the StakeAddressState. Depends on SPDD from #7

[sandtreader]

Moved back to in-progress because although main structure is landed on main, we are still working to fix the reward calculation results.