feat: Batch value sync requests

[hvanz]

Currently fetching values from peers happens one block at a time. We want an improved version of the sync protocol to request a range of values.

[bastienfaivre]

sequenceDiagram
    box Syncing Node
        participant A1 as Application
        participant C1 as Consensus
        participant S1 as Sync
    end

    box Helping Peer
        participant S2 as Sync
        participant C2 as Consensus
        participant A2 as Application
    end

    Note over A1,C1: Normal Operation
    A1->>C1: StartHeight(height)
    C1->>S1: StartedHeight(height)
    S1->>S2: BroadcastStatus(height)

    Note over A1,C1: Node Falls Behind
    S2->>S1: Status(peer_height)
    S1->>S2: GetValuesRequest(from_height, to_height, max_size)
    S2->>A2: GetDecidedValues(from_height, to_height)
    A2->>S2: DecidedValues(ord_map(height -> (value, certificate)))
    S2->>S1: ValuesResponse(ord_map(height -> (value, certificate)))
    S2->>C1: ValuesResponse(ord_map(height -> (value, certificate)))

    Note over A1,S1: For each height's value, do:

    C1->>C1: ProcessCommitCertificate(certificate)
    C1->>A1: ProcessValue(height, value)
    A1->>C1: ProposedValue(height, value, validity)
    C1->>A1: Decided(height, certificate)
    C1->>S1: Decided(height)

    Note over A1,S1: End for

Loading

Issues:

• Suppose we request values for heights 1-20 from another peer, and in the response, all values are correct, except the one for height 1 (or any technically any height <20). This prevents the consensus actor from making progress. So, we would need to request this value again (potentially from another peer). The problem is that the sync actor has no way of knowing whether the remaining values (2-20) are correct. It would be highly inefficient to re-request all these values again. Moreover, are these other values cached by the consensus actor (replay values)?
• The last point is crucial since batch processing might accentuate the unintuitive behaviour observed with Snapchain nodes: what should the sync actor do while waiting for a sync response to be fully processed?

[adizere]

Suppose we request values for heights 1-20 from another peer, and in the response, all values are correct, except the one for height 1 (or any technically any height <20). This prevents the consensus actor from making progress. So, we would need to request this value again (potentially from another peer). The problem is that the sync actor has no way of knowing whether the remaining values (2-20) are correct. It would be highly inefficient to re-request all these values again. Moreover, are these other values cached by the consensus actor (replay values)?

Answer from @sergio-mena

If the first piece is wrong, the rest of 19 are irrelevant. The sender is faulty, so throw them away.
We're doing batching for optimization.