Replace O(n²) sync algorithms with HashMap lookups

[schjonhaug]

Parent: #236 (Phase 1)

Problem

CPFP detection and RBF conflict detection in sync.rs use .find() inside loops over all BDK transactions, resulting in O(n²) complexity. At 50K+ transactions, sync takes minutes instead of milliseconds.

What to change

CPFP detection

Replace:

all_bdk_txs.iter().find(|tx| tx.txid.to_string() == *txid)

With a pre-built HashMap<String, &CanonicalTx> keyed by txid, built once before the loop.

RBF conflict detection

Same pattern — the nested loop searches all_bdk_txs linearly for each conflicted and canonical txid. Build the HashMap once, do O(1) lookups.

Expected result

• O(N) total instead of O(N²)
• 50K tx sync: <100ms instead of minutes
• Memory: one HashMap allocation (~50K entries) instead of repeated linear scans

Key files

• backend/src/sync.rs — CPFP detection and RBF conflict detection

[schjonhaug]

Completed in PR #245.