feat(extract): one normalisation door removes U+00AD before the guard

Operator decision 2026-09-18, and the whole of it: okf removes the soft
hyphen, the guard is untouched, and the number is BOOKED.

`normalise_extracted(text) -> (text, removed)` in `extract`, applied
once at the end of `extract_document`, so both entry points and both
sides of a plan's `text_sha256` see one string. `_pdf_units` measures
its page offsets through the same door, because that table is
CHARACTER offsets rebuilt from the pages while the text it indexes
comes back normalised -- two readings of one document, and a table
built against the other names the wrong page with full confidence.
Removing a character removes no newline, so every line-indexed rule
downstream is unmoved by construction.

`ExtractedDocument.soft_hyphens` carries the count out;
`InboxResult.normalised` carries it per document; `account_run` reads
it off the RUN rather than recounting the source, because a second
count would be a second reader. It reaches the accounting JSON as
`normalised_soft_hyphen` at both levels and `log.md` as a
`**Normalisation**` bullet naming the count, the documents touched and
that no other character is moved.

EXPOSURE, measured and not assumed: 0 of the 78 readable documents of
the K2 reference corpus carry U+00AD or any of the four real
zero-width characters (the 8 unreadable ones raise before extraction),
0 in the pinned K2 bundle's concept bodies, and 0 across
`tests/fixtures`, `examples`, `skills`, `docs`, `src`, README and
CHANGELOG. The door cannot have moved a byte anyone here has measured.

Suite 2179 passed, 1 skipped; ruff and `mypy --strict src/` clean.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

src/llm_ingestion_okf/accounting.py

@@ -732,6 +732,12 @@ class DocumentAccount:
     counts: dict[str, int]
     fates: dict[str, Fate]
     error: str | None = None
+    #: How many U+00AD the normalisation door removed from this document's
+    #: text before the persist gate saw it (operator decision 2026-09-18).
+    #: Booked rather than silently applied: a door that changes a source's
+    #: bytes and says nothing is the same class of absence this whole module
+    #: exists to close.
+    normalised_soft_hyphen: int = 0
 
     @property
     def unaccounted(self) -> dict[str, int]:
@@ -758,6 +764,7 @@ class DocumentAccount:
             "source_file": self.source_file,
             "status": self.status,
             "code": self.code,
+            "normalised_soft_hyphen": self.normalised_soft_hyphen,
             "inventory": dict(self.counts),
             "fates": {kind: self.fates[kind].to_json() for kind in self.counts},
             "unaccounted": self.unaccounted,
@@ -806,6 +813,11 @@ class Accounting:
         """
         return sum(1 for d in self.documents if d.status == REJECTED)
 
+    @property
+    def normalised_soft_hyphen(self) -> int:
+        """Soft hyphens the normalisation door removed across the whole run."""
+        return sum(d.normalised_soft_hyphen for d in self.documents)
+
     @property
     def images_found(self) -> int:
         return sum(d.counts.get("image", 0) for d in self.documents)
@@ -820,6 +832,7 @@ class Accounting:
             "unaccounted": self.unaccounted,
             "double_booked": self.double_booked,
             "refused": self.refused,
+            "normalised_soft_hyphen": self.normalised_soft_hyphen,
             "documents": [d.to_json() for d in self.documents],
             "files": [f.to_json() for f in self.files],
         }
@@ -831,6 +844,12 @@ class Accounting:
             f"{self.unaccounted} unaccounted, {self.double_booked} double-booked; "
             f"{self.refused} of {len(self.documents)} document(s) refused whole."
         ]
+        touched = sum(1 for d in self.documents if d.normalised_soft_hyphen)
+        lines.append(
+            f"* **Normalisation**: {self.normalised_soft_hyphen} soft hyphen(s) (U+00AD) "
+            f"removed from {touched} of {len(self.documents)} document(s) before the persist "
+            "gate. No other character is touched."
+        )
         for doc in self.documents:
             if doc.status == REJECTED:
                 lines.append(
@@ -986,6 +1005,12 @@ def account_run(inbox: Path, walked: Sequence[Path], result: InboxResult) -> Acc
             documents.append(
                 DocumentAccount(name, "", None, inv.counts(), {k: Fate() for k in inv.counts()})
             )
+    # Read off the run, never recounted from the source: a second count would
+    # be a second reader, and the number this publishes has to be the number
+    # the normalisation door acted on.
+    removed = {item.source_file: item.soft_hyphens for item in result.normalised}
+    for document in documents:
+        document.normalised_soft_hyphen = removed.get(document.source_file, 0)
     return Accounting(documents, files)