Respect user batch's reference block

[Mirko-von-Leipzig]

First, we are not respecting all of the batch parameters. Specifically, a user-submitted batch as a specific reference block number, but IIUC, we are ignoring it here. So, a user-submitted batch is different from a batch that is built by the node because the node always picks the latest block to "anchor" the batch against, while the user may have a different "anchor" block. In the future, we may have additional "user-defined" properties for a batch that may not be inferred from just the transaction list.

[giwaov]

I can take this one if it is available. Could you assign it to me before I start?

[AlgofootPrint]

Hi @Mirko-von-Leipzig, I’d like to work on this if it’s available.

I traced the current path and it looks like the batch-level reference is lost when the block producer handles SubmitProvenBatch: RPC verifies the ProvenBatch, but the block producer only extracts transactions from the ProposedBatch and calls Mempool::add_user_batch(&txs). From there, SelectedBatch only carries transactions/account updates, so later batch building derives inputs from transaction references instead of preserving the user-submitted batch reference.

Proposed scope:

• preserve the user-submitted batch reference metadata through add_user_batch / SelectedBatch,
• make batch input selection respect that reference for user batches,
• add focused mempool/batch-builder coverage showing a user batch keeps its submitted reference block,
• keep conventional node-built batches unchanged.

Could you assign this to me?