fix: prevent endless retry when CL times out validating large payloads

[ZZZ0h]

This PR addresses the endless retry issue caused by CL (Consensus Layer) timeout when validating large payloads in PBFT consensus model. The problem occurs because:

1. Hash mismatch after execution: The block hash changes after execution when gas_used is updated from 0 to actual value, causing standard hash-based lookup to fail on retry.
2. Infinite retry on invalid payload: The CL may repeatedly send the same invalid payload if it times out before receiving the INVALID response.

Changes:

1. Add executed block lookup by (number, parent_hash):

   • Add executed_block_by_number_and_parent method to TreeState
   • Before executing a payload, check if it has already been executed using (block_number, parent_hash) pair
   • Validate payload structure without re-executing and return previous execution result
   • This avoids redundant EVM execution for the same payload on timeout retry

2. Add fast reject for known invalid blocks:

   • Early check in newPayload handler for blocks already marked as invalid
   • Immediately return INVALID response without re-executing signature recovery or EVM
   • Prevents infinite retry loop for invalid payloads

These changes significantly improve system stability and resource utilization under high-pressure testing with large blocks.