test(store): a write-off whose notice is another entry's receipt is refused before posting

The defect found merging S7 and P6: under P6's dedup the notice's
bytes resolve to the record already on file, and the one-receipt-one-
entry rule refused it only after the write-off had journaled its
entry. The store now checks the deduplicated record before the journal
line (`6c26160e`'s companion fix in `write_off_invoice`); this test
pins it: the refusal names the record and its entry, the head and the
journal are unchanged, no entry posts, the invoice stays open, and the
books verify. Also formats the B-27 test `ci-fixes` brought in.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HS8WZziDnyxunC23MUxrLp

crates/ledger-store/tests/format.rs

@@ -1574,7 +1574,8 @@ fn an_unknown_convention_kind_is_newer_for_the_writer_and_the_viewer_alike() {
     }
     let store = ledger_store::Store::open(dir);
     let snap = ledger_store::Snapshot::load(dir);
-    let newer = |e: &ledger_store::StoreError| matches!(e, ledger_store::StoreError::NewerJournal { .. });
+    let newer =
+        |e: &ledger_store::StoreError| matches!(e, ledger_store::StoreError::NewerJournal { .. });
     match (store, snap) {
         (Err(a), Err(b)) => assert!(
             newer(&a) && newer(&b),