BEEFY: generate historical proofs

[svyatonik]

Polkadot companion: paritytech/polkadot#5334

This PR adds new MMR RPC method - mmr_generateHistoricalProof. It is quite similar to the existing mmr_generateProof method, but it has additional argument - number of MMR leaves. It may be used to generate proof using historical MMR state even using latest block state, where there are more leaves.

Where it is useful:

1. BEEFY light client needs to import explicit commitment + leaf + leaf proof for blocks that are changing validator set. If the chain has progressed since such block, the state of validator-set-change-block may have already been discarded (unless you're using archive node, which we want to avoid) and the call of mmr_generateProof(use-state-at-validator-set-change-block) will obviously fail. We may use mmr_generateHistoricalProof(use-state-at-best-block, use-mmr-state-at-validator-set-change-block) instead - this will generate us required proof;
2. if block B is known to the light client, then we can't simply generate some MMR-related proof using MMR state at block B+10. So e.g. to prove some storage (read: deliver message), we need to generate proof using MMR state at block B. If state is discarded, we'll need to import B+10 first => extra bridge cost.

The leaves_count argument can easily be computed by the client by assuming that there's exactly one leaf per every block starting from some beginning block (where MMR pallet has been deployed). We may read number of MMR leaves at the best_block and then compute leaves_count_at_historical_block as number_of_historical_block - (best_block_number - leaves_count_at_best_block).

[acatangiu]

Would it make sense to directly add this with support for multiple leaves at once? (see #10635)

Would the light client benefit for generating historical proofs in a batch?

[svyatonik]

Yeah - I think it makes sense to add support for proving several leaves, even though we'll only be using this API to prove single leaf now.

[svyatonik]

Then let's probably defer this PR until #10635 is merged to reuse primitives from there && etc

[tomusdrw]

IMHO it makes sense to add an explicit error in case leaves_count > Self::mmr_leaves()

[stale]

Hey, is anyone still working on this? Due to the inactivity this issue has been automatically marked as stale. It will be closed if no further activity occurs. Thank you for your contributions.