observations_gate: normalise emphasis before matching a marker (R-419 follow-up)

My first fix was too strict. It anchored a marker to a line start or a bare '. ', which misses the
commonest real shape - a bolded sentence followed by a bolded marker: '...them.** **FILED: R-427**'.

CAUGHT BY THE GATE CONVICTING THE VERY REPORT THAT DOCUMENTS IT, on two of its own observations.
That is the both-directions check working: a gate that rejects the decoy AND the genuine article is
worse than the hole it replaced.

Code spans are stripped FIRST and that order matters - a marker inside backticks is being talked
about, never used, and removing it is what makes the R-419 decoy fail. Emphasis is stripped second so
'**FILED: R-1**' and 'FILED: R-1' are the same thing to the anchor.

Decoy suite re-run: the R-419 decoy and a backticked-only mention are still REFUSED; both genuine
marker shapes pass.

scripts/observations_gate.py

@@ -89,15 +89,29 @@ ITEM_RE = re.compile(r"^\s*(\d+)[.)]\s+(.*)$")
 # where a real marker is written and where a mention inside prose never is. `**FILED: R-417**` at
 # the end of a sentence is the common real shape, so a marker is also accepted after a sentence
 # boundary — but never mid-sentence and never inside backticks.
-_MARK = r"(?:^|(?<=[.!?)]\s)|(?<=[.!?)]\s\s))[\s>*_\-]*"
-FILED_RE = re.compile(_MARK + r"\*{0,2}FILED:\*{0,2}\s*(R-\d+)",
-                      re.IGNORECASE | re.MULTILINE)
-NOT_A_FINDING_RE = re.compile(_MARK + r"\*{0,2}NOT-A-FINDING:\*{0,2}\s*(\S.*)$",
-                              re.IGNORECASE | re.MULTILINE)
+# Emphasis is NORMALISED AWAY before matching (see normalise_for_markers), so the anchor only has to
+# describe where a marker sits in a SENTENCE: at the start of a line, or after a sentence ends. The
+# common real shape is a bolded sentence followed by a bolded marker — `...them.** **FILED: R-427**` —
+# and an anchor that did not allow the emphasis run between them rejected genuine markers. That was
+# caught by this gate convicting the very report that documents it.
+_MARK = r"(?:^|(?<=[.!?)]\s))\s*"
+FILED_RE = re.compile(_MARK + r"FILED:\s*(R-\d+)", re.IGNORECASE | re.MULTILINE)
+NOT_A_FINDING_RE = re.compile(_MARK + r"NOT-A-FINDING:\s*(\S.*)$", re.IGNORECASE | re.MULTILINE)
 
 # A marker written inside backticks is being TALKED ABOUT, never used. Strip inline code spans
 # before matching — this is what makes the R-419 decoy fail.
 CODE_SPAN_RE = re.compile(r"`[^`]*`")
+EMPHASIS_RE = re.compile(r"[*_]+")
+
+
+def normalise_for_markers(body):
+    """Strip inline code spans, then emphasis, before looking for a marker.
+
+    CODE SPANS GO FIRST AND THAT ORDER MATTERS: a marker inside backticks is being TALKED ABOUT,
+    never used, and removing it is what makes the R-419 decoy fail. Emphasis goes second so that
+    `**FILED: R-427**` and `FILED: R-427` are the same thing to the anchor below.
+    """
+    return EMPHASIS_RE.sub("", CODE_SPAN_RE.sub("", body))
 
 REGISTERS = [
     os.path.join(ROOT, "documentation", "backlog", "OPEN-ITEMS.md"),
@@ -204,7 +218,7 @@ def main():
     convictions = []
     satisfied = []
     for num, body in items:
-        body = CODE_SPAN_RE.sub("", body)   # R-419: a marker inside backticks is a mention
+        body = normalise_for_markers(body)  # R-419
         filed = FILED_RE.findall(body)
         declared = NOT_A_FINDING_RE.findall(body)
         first_line = body.split("\n")[0].strip()