No DA verification for sequencer commitments in batch proof

[eyusufatik]

Our current batch proof logic is defined as following:

1. The batch proof circuit takes in is a list of sequencer commitments found on DA as DA level blobs.
2. The sequencer commitments commit to correct hashes for a given list of L2 block ranges.
3. When the L2 block ranges in sequencer commitments are run, the state transitions from state root A to state root B

This feature is interested in getting rid of 1. Since the batch proof is not recursive, and it has no rules around supplying a valid DA header chain, theoretically, the batch prover can make up a DA block and supply that to the batch proof circuit. However, it can’t make a false sequencer commitment, as the commitment is signed by the sequencer. This means, the DA inclusion verification done for the sequencer commitment is useless.

The proposed solution is to remove DA verification altogether from the batch proof circuit, and accept the sequencer commitments as an ordinary input field. The new batch proof circuit would output the hash found in the sequencer commitment, which is then read by the light client proof circuit to check against an MMR tree of sequencer commitments.

New logic then goes as:

1. The batch proof circuit takes in a list of sequencer commitments. These are now in the form of the struct SequencerCommitment not DA level blobs.
2. The sequencer commitments commit to correct hashes for a given list of L2 block ranges.
3. When the L2 block ranges in sequencer commitments are run, the state transitions from state root A to state root B.
4. The batch proof includes the hashes extracted from sequencer commitments in its output.
5. When the light client prover processes a batch proof, it extracts the sequencer commitment hashes from the batch proof output, and verifies these hashes against its sequencer commitment MMR.