feat(rpc)!: Update Emitter::mempool to support evicted_at

[LagginTimes]

Fixes #1740

Description

Work for this began as part of #1811, based on this comment.

Emitter::mempool now returns a MempoolEvent which provides the data for tracking evicted_at:

pub struct MempoolEvent {
    /// Unemitted transactions or transactions with ancestors that are unseen by the receiver.
    ///
    /// To understand the second condition, consider a receiver which filters transactions based on
    /// whether it alters the UTXO set of tracked script pubkeys. If an emitted mempool transaction
    /// spends a tracked UTXO which is confirmed at height `h`, but the receiver has only seen up to
    /// block of height `h-1`, we want to re-emit this transaction until the receiver has seen the
    /// block at height `h`.
    pub new_txs: Vec<(Transaction, u64)>,

    /// [`Txid`]s of all transactions that have been evicted from mempool.
    pub evicted_txids: HashSet<Txid>,

    /// The latest timestamp of when a transaction entered the mempool.
    ///
    /// This is useful for setting the timestamp for evicted transactions.
    pub latest_update_time: u64,
}

Changelog notice

• Change Emitter::mempool to return MempoolEvents which contain mempool-eviction data.
• Change Emitter::client to have more relaxed generic bounds. C: Deref, C::Target: RpcApi are the new bounds.
• Add conversion impls for CanonicalTx to Txid/Arc<Transaction>.
• Add ChainPosition::is_unconfirmed method.

Checklists

All Submissions:

• I've signed all my commits
• I followed the contribution guidelines
• I ran cargo fmt and cargo clippy before committing

New Features:

• I've added tests for the new feature
• I've added docs for the new feature

Bugfixes:

• This pull request breaks the existing API
• I've added tests to reproduce the issue which are now passing
• I'm linking the issue being fixed by this PR

[evanlinjin]

Instead of rebasing, it's probably best to git switch -c <new-branch> from #1839 and cherry-pick on the new commits.

[evanlinjin]

Documentation is outdated, but I understand this is WIP.

[evanlinjin]

I haven't finished reviewing yet. Some thoughts:

1. We probably need something like insert_relevant_evicted_ats(&mut self, evicted_txs: impl IntoIterator<Item = (Txid, u64)>) on TxGraph for convenience.

2. The old best-effort-avoid-reemitting-mempool-txs logic might actually be replaceable by caching Arc<Transaction> in expected_mempool_txids.

   prior_mempool_txs: HashMap<Txid, Arc<Transaction>>,

   The idea is we always emit the whole mempool but it is still memory-efficient.

For 2., I guess we can tackle that in a separate PR.

[evanlinjin]

The logic looks solid.

There’s room for improvement in documentation clarity across tests, inline code comments, and public doc comments:

• Code comments: Focus on the why behind the logic, rather than describing what the code is doing—since that’s usually self-evident from the code itself.
• Tests: For larger tests, consider adding a summary doc comment at the top. This helps readers quickly understand what the test is verifying without having to trace through each line and reconstruct the intent.
• Public doc comments: Prioritize what’s important for the caller to know—avoid going into implementation details unless they directly impact usage.

[evanlinjin]

Looks good, just a final request.

[evanlinjin]

Apologies for the incremental reviews — it's been eye-opening for me as well how tricky the Emitter is to reason about. Let's aim to simplify it in follow-up PRs.

One thing I’ve come to realize: the next_header method is inconsistent with the eviction-detection logic we're introducing via next_block and mempool. I suggest removing next_header entirely — everything it does can already be achieved through next_block.

[evanlinjin]

self-ACK e1c53d7

[ValuedMammal]

@LagginTimes Can you rebase on master and also squash 6e1178c into the previous commit?

[ValuedMammal]

ACK 8513d83

[ValuedMammal]

Thanks @LagginTimes

I didn't mean to become author on 28ef7c9 😅