feat: euclid phase-2 contract changes

[zimpha]

This PR contains all changes for the planned EuclidV2 upgrade. Specifically,

• Reduce commit cost: we reduce the cost to commit batches significantly by
  1. move calldata (chunks and block contexts) to blob -> no notion of chunks anymore
  2. chaining of L1 messages with MessageQueueV2: instead of iterating through L1 messages and computing the hash at commit time, L1 messages are chained with a rolling hash. so that the last hash represents a commitment to the entire queue.
  3. move point evaluation to circuit
  4. commit multiple batches in a single transaction
• Enforced transactions: transactions are appended to MessageQueueV2 together with the time they have been submitted to L1. If not included/finalized for maxDelayMessageQueue, the operator can not finalize any more batches until including those messages. If it continues to censor messages permissionless mode will eventually activate.
• Enforced/permissionless batches: if no batch has been finalized for maxDelayEnterEnforcedMode, batch submission becomes permissionless. Anyone can submit a batch together with a proof.
• Deprecate Clique: we introduce a SystemConfig contract that contains the address of the currently valid signer for L2 blocks. L2 nodes can read and will only accept L2 blocks signed by the given address.

[jonastheis]

Some questions:

• maybe I missed it but do we ensure that all messages on MessageQueueV1 are finalized before we allow finalization of MessageQueueV2 messages?
• for enforced transactions: don't we need to check that all messages on the queue that are older than a certain threshold are included when finalizing? otherwise we don't actually force the sequencer to include them. it could keep censoring forever and permissionless mode would not activate either.

[zimpha]

Some questions:

• maybe I missed it but do we ensure that all messages on MessageQueueV1 are finalized before we allow finalization of MessageQueueV2 messages?
• for enforced transactions: don't we need to check that all messages on the queue that are older than a certain threshold are included when finalizing? otherwise we don't actually force the sequencer to include them. it could keep censoring forever and permissionless mode would not activate either.

1. good catch, I can add the check
2. probably I forget to add the check, since this code is not added in previous PR.

The base branch was changed.