qa: strip 'all/there/know/everything' from retrieval queries; bench gains deflection rate

Two paired changes:

(1) Retrieval template-phrase stopwords. Bench finding 2026-04-30:
'tell me all there is to know about york england?' missed the
en.wikipedia.org/wiki/York article entirely — corpus has it but the
FTS5 AND-mode MATCH expanded to "all" AND "there" AND "know" AND
"york" AND "england", which filters to only docs containing every
template word. Articles like "Charles I of England" happen to mention
"all" + "there" + "know" + "england" enough to win; the bare "York"
stub doesn't, so it falls out of the top-K.

Added `all there know everything anything something` to both
_FTS5_STOPWORDS (search/fts5.py) and _TITLE_STOPWORDS (qa/query.py)
so they stay in sync. After the change "York" article ranks #1 on
both 'tell me all there is to know about' and 'tell me everything
about' phrasings.

These are retrieval-time stopwords only — they don't enter the
cache_key or governance_policy_hash (cache_key hashes the original
question text via question_hash). Existing cached records remain
valid; new queries get better retrieval.

(2) bench/qa_sweep.py: per-row deflection_kind + subject_anchor +
subject_in_answer (from qa.inspect.diagnose_deflection); summary
table grows a 'deflections' column tracking STRICT/HYBRID rows
where the question's subject anchor is missing from the answer.
Surfaces the Mars-BDFL pattern at aggregate scale.

aborist/qa/query.py

@@ -109,6 +109,7 @@ _TITLE_STOPWORDS = frozenset(
     it we they me him her us them do does did have has had can could should
     would will may might
     tell show describe explain summarize say give list find make please
+    all there know everything anything something
     """.split()
 )
 

aborist/search/fts5.py

@@ -26,6 +26,7 @@ _FTS5_STOPWORDS = frozenset(
     do does did have has had can could should would will may might
     tell show describe explain summarize say give list find make
     please
+    all there know everything anything something
     """.split()
 )
 

bench/qa_sweep.py

@@ -100,6 +100,14 @@ def _run_one(
         err = f"{type(e).__name__}: {e}"
     elapsed_s = round(time.monotonic() - t0, 2)
 
+    # Deflection sidecar — sidecar-only signal, never feeds the
+    # verifier. Surfaces the Mars-BDFL pattern (STRICT but answer
+    # never mentions the question's subject) at bench-aggregate
+    # scale so a creeping "model deflects rather than refuses"
+    # regression is legible across runs.
+    from aborist.qa.inspect import diagnose_deflection
+    deflection = diagnose_deflection(question, result.get("answer_text") or "")
+
     audit_mode = result.get("audit_mode")
     return {
         "question": question,
@@ -117,6 +125,9 @@ def _run_one(
         "cache_key": (result.get("cache_key") or "")[:12],
         "n_sources": len(result.get("sources") or []),
         "elapsed_s": elapsed_s,
+        "deflection_kind": deflection["kind"],
+        "subject_anchor": deflection["subject_anchor"],
+        "subject_in_answer": deflection["subject_in_answer"],
         "error": err,
     }
 
@@ -134,6 +145,7 @@ def _summarize(rows: list[dict]) -> dict:
             "errors": 0,
             "ratio_sum": 0.0,
             "latency_sum": 0.0,
+            "deflections": 0,
         })
         b["n"] += 1
         if r["error"] or r["status"] == "error":
@@ -142,6 +154,13 @@ def _summarize(rows: list[dict]) -> dict:
             b[r["audit_mode"]] += 1
         b["ratio_sum"] += r["ratio"] or 0.0
         b["latency_sum"] += r["elapsed_s"] or 0.0
+        # Deflection rate: STRICT/HYBRID with subject-anchor missing
+        # from answer. UNGROUNDED + deflection isn't interesting (no
+        # answer to deflect with), and no_question_tokens is vacuous.
+        if r.get("deflection_kind") == "deflection" and r["audit_mode"] in (
+            "STRICT", "HYBRID"
+        ):
+            b["deflections"] += 1
     return by_mode
 
 
@@ -190,8 +209,8 @@ def _render_markdown(
     lines.append("")
     lines.append("## summary")
     lines.append("")
-    lines.append("| mode | runs | STRICT | HYBRID | UNGROUNDED | err | strict-rate | mean ratio | mean latency |")
-    lines.append("|------|------|--------|--------|------------|-----|-------------|-----------|--------------|")
+    lines.append("| mode | runs | STRICT | HYBRID | UNGROUNDED | err | strict-rate | mean ratio | mean latency | deflections |")
+    lines.append("|------|------|--------|--------|------------|-----|-------------|-----------|--------------|-------------|")
     for mode in modes:
         b = summary.get(mode)
         if not b:
@@ -200,10 +219,12 @@ def _render_markdown(
         mean_ratio = b["ratio_sum"] / n
         mean_lat = b["latency_sum"] / n
         strict_rate = b["STRICT"] / n
+        deflections = b.get("deflections", 0)
         lines.append(
             f"| {mode} | {b['n']} | {b['STRICT']} | {b['HYBRID']} | "
             f"{b['UNGROUNDED']} | {b['errors']} | "
-            f"{strict_rate:.2f} | {mean_ratio:.3f} | {mean_lat:.1f}s |"
+            f"{strict_rate:.2f} | {mean_ratio:.3f} | {mean_lat:.1f}s | "
+            f"{deflections}/{b['n']} |"
         )
     lines.append("")
     lines.append("## per question")