op-node: Span Batch Type, Encoding, and Decoding

[ImTei]

Contexts

This PR contains the type, encoding, and decoding code for Span Batch as defined specs.

Please refer to the Encoding Optimization section in Design Docs for the rationale behind the current encoding specs.

This PR has a dependency on the Hardfork Configuration PR.

Changes

• Added the following types:
  • Batch interface
  • RawSpanBatch
  • SpanBatchElement
  • SpanBatch
  • SpanBatchBuilder
• Changed the following types:
  • Renamed BatchV1 to SingularBatch.
  • BatchData
• ^ Please refer to the Type section of Implementation Design Docs for details.
• Added unit tests.
  • ^ Please refer to the Data Structure category in the Test List Sheet for each test’s details.

[mergify]

Hey @ImTei! This PR has merge conflicts. Please fix them before continuing review.

[protolambda]

Adding a do-not-merge label to give others the chance to review before auto-merge by mergify kicks in.

[protolambda]

Looks good! There are some stylistic but breaking things I think we can do in a follow-up PR, after the PR stack is wrapped up.

[ImTei]

applied simple changes from code reviews and rebased to the latest develop branch.

Other suggestions will be applied in the follow-up PR!

[protolambda]

Follow-ups to implement in later PR:

• spanBatchTxs has unused EncodeRLP and DecodeRLP. The marshal/unmarshal-binary functions are enough.
• Some encoding/decoding tests in span_batch_tx_test.go randomize the test data and assert
  if the tx types all occurred at the end. Instead, we can make sub-tests with t.Run() for each tx-type,
  and randomize the same amount of txs per type.
• The BatchData wrapper around the typed batch shouldn't have an explicit byte field for the type:
  it can already be inferred from the contained batch data. See types.Transaction in geth as example.
• decodeTyped can be simplified, and the switch-statement should have be more consistent (not mix constants and data[0])
• SingularBatchType = iota is neat, but being explict about protocol constants is better. iota is only really meant for internal enumerations that you can flexibly add/remove to.
• Review usage of require vs assert in batch tests
• parentCheck []byte and l1OriginCheck []byte should be [20]byte types

And opened #7615 to fix a blocking issue: we must not allow span-batches to be decoded successfully before the HF, or we diverge from the actual mainnet specs.