Align block structure with yellow paper

[LHerskind]

The block as defined currently is very hard to follow. It uses different names all over the place to refer to the same things. We should use structs when possible, and be more consistent in the naming.

For example, the L2Block don't explicitly define the block header, so the BlockHeader in typescript is not consistent with the actual block. It is only the "after" half of it.

Instead, we must update it to match what is described in the yellow-paper.

Tasks

Preview Give feedback

1. Rename the blockstree to archive #3572
   +0
   
2. Make noir-circuits independent of aztec-nr #3589
   +0
   
3. Reduce duplication across aztec-nr and protocol-circuits #3592
   +0
   
4. Add StateDiffHints and update BaseRollupInputs #3916
   +0
   
5. Create and use PartialStateReference and StateReference noir structs #3820
   +0
   
6. Re-enable the old state check in Rollup and L1Publisher #3936
   +0
7. Update L2 block encoding to use Header #4010
   +0
   
8. Update Rollup.process function to accept params according to yellow paper #4011
   +0
9. Restructure the API's to fetch "header" instead of individual parts and then building anyway
10. Update L2 block structure to include Body struct #3868
    +0
    
11. refactor: Update the TxsHash + add TxEffect #3938
    +0
    
12. Update BaseOrMergeRollupPublicInputs #3849
    S-blocked +1
    
13. Update L1 state root definition (pedersen or sha)
14. Update GlobalVariables #3824
    12 of 12
    +0
    
15. Replace all the use of the old BlockHeader struct with the new Header #3937
    +0
    
16. Update block hash computation #3941
    +0
    
17. Re-enable public inputs hash and body hash (originally calldata hash) checks #4031
    +0
    
18. Unify ordering of values in Header serialization and ordering of its constructor params #4045
    +0
    
19. Use new header in get_block_header #4134
    +0
    
20. Remove outdated kernel vk tree stuff #3441
    +0
21. Update how tx effects are encoded in a block #4720
    +0

[LHerskind]

The blocksTreeRoot also is a bad name, hence it is not actually a tree of blocks, but a tree of StateContext from the above definitions (almost).

It should either be renamed, or consist of commitments to the actual blocks.

Our apis which are mostly used for building the header should also be altered to return a StateReference instead and header building should be more easily exposed to the PXE.

[LHerskind]

From discussion earlier, we came up with a few structs and hopefully some better naming:

// Base rollup only change part of the state
interface PartialStateReference {
    noteHashTreeRoot: Snapshot;
    nullifierTreeRoot: Snapshot;
    contractTreeRoot: Snapshot;
    publicDataTreeRoot: Fr;
};

// StateRoot = H(StateReference)
// Tree of StateRoots = Archive
interface StateReference {
    ...partialStateReference;
    privateKernelVkTreeRoot: Snapshot;
    l1ToL2MessageTreeRoot: Snapshot; 
    archiveRoot: Fr; // Before insertion of this leaf
    globalVariablesHash: H(GlobalVariables);
};

interface CombinedConstantData {
    historical_state: StateReference; // old header
    context: TxContext;
};

interface BlockHeader {
    parent: StateRoot;
    state: StateReference;
    globalVariables: GlobalVariables;
}

interface Block {
    header: BlockHeader;
    content: Content;
}

Further, there are some cleanup to do regarding global variables which have multiple variations in noir which makes it hard to follow.

[LHerskind]

As mentioned in #3564 it will also be useful to expose the block number as part of the state reference to make historic access easier. Practically, this could be done by not providing a globalVariablesHash but actually provide the global variables, that way we can just read them directly, and as we are not in the getting all the juice out of the circuits state, this seems like a valid route to make it easier to understand what is going on and build.

[LHerskind]

Closing this as its subtasks have been addressed.