qa/tests: cover capacity metrics + retrieval-keywords across all 3 layers

Three layers, 14 new tests, full suite 611 passed (was 597):

UNIT — tests/test_query.py
  test_query_returns_prompt_chars_breakdown
    Asserts the result dict's prompt_chars carries exactly the five
    expected keys and messages_total equals sum of message contents
    the StubClient saw.
  test_query_answer_chars_matches_answer_text
    answer_chars == len(answer_text) — drift check.
  test_query_cache_hit_also_returns_capacity_metrics
    Cache-hit path populates prompt_chars + answer_chars (operators
    inspecting cached records still want the breakdown).
  test_query_evidence_chars_grows_with_topk
    Sanity: more sources / larger budget → more evidence chars
    (the metric tracks actual context build, not a stale constant).

INTEGRATION — tests/test_query.py (retrieval_keywords)
  test_retrieval_keywords_does_not_alter_question_to_llm
    Keywords don't appear in the LLM-facing question segment;
    system prompt unchanged across runs. Pins the substrate
    contract: keywords are FTS5/title-filter-only.
  test_retrieval_keywords_changes_retrieved_sources
    Different keyword sets surface different docs (the actual
    user-visible behavior).

UNIT — tests/test_cli_render.py
  test_render_shows_capacity_line_when_prompt_chars_present
    Capacity line appears with messages_total + breakdown when
    prompt_chars is in the result dict.
  test_render_omits_capacity_line_on_legacy_results_without_prompt_chars
    Backwards-compat: legacy results render cleanly without the
    capacity line — no KeyError, no '0 chars' noise.
  test_render_capacity_thousand_separators
    61,550 not 61550 — operator legibility on daily renders.

UNIT/INTEGRATION — tests/test_bench_qa_sweep.py (NEW FILE)
  Imports bench/qa_sweep.py via importlib.util so the module's
  not in the Python path doesn't matter. Five tests:
    - _summarize counts verdicts by mode
    - deflections counted only on STRICT/HYBRID rows (not UNGROUNDED)
    - rendered markdown has the headline summary + size buckets
    - size buckets correctly stratify strict-rate by prompt_chars_total
    - empty buckets are skipped (no '0 runs' noise)

FUNCTIONAL — live verification (no automated test, manual)
  `make query Q="what is the capital of france?" BURN=1` confirmed
  in commit f927298 to render the capacity one-liner under the
  source list. Documented in that commit's body.

Also corrected the docstring on query()'s `retrieval_keywords` to
reflect that keywords don't enter cache_key DIRECTLY but do change
context_root + conversation_hash via source selection — so the same
question with different keywords lands under different cache_keys
(legitimately, since the LLM saw different contexts).

tests/test_cli_render.py

@@ -146,3 +146,60 @@ def test_json_mode_also_uses_unicode():
     s = json.dumps(payload, ensure_ascii=False)
     assert "Pokémon" in s
     assert "\\u00e9" not in s
+
+
+# ---------------------------------------------------------------------------
+# capacity metrics — human render surfaces prompt_chars breakdown
+# ---------------------------------------------------------------------------
+
+
+def test_render_shows_capacity_line_when_prompt_chars_present():
+    """When the result carries `prompt_chars`, the human render
+    emits a one-liner summarizing prompt size + answer size so an
+    operator can tell at a glance whether STRICT came from a small
+    prompt or a context-stuffed one."""
+    r = _result(
+        prompt_chars={
+            "system_prompt": 1292,
+            "grounding_reminder": 342,
+            "user_question": 30,
+            "evidence_or_context": 20304,
+            "messages_total": 21996,
+        },
+        answer_chars=247,
+    )
+    out = _render_query_human(r, "what is the capital of france?")
+    assert "capacity:" in out
+    assert "21,996 chars" in out  # messages_total
+    assert "sys 1,292" in out
+    assert "evidence 20,304" in out
+    assert "answer 247 chars" in out
+
+
+def test_render_omits_capacity_line_on_legacy_results_without_prompt_chars():
+    """Backwards-compat: results that pre-date capacity metrics (no
+    prompt_chars key) render without the capacity line — no KeyError,
+    no awkward 'capacity: 0 chars' noise."""
+    r = _result()  # no prompt_chars
+    out = _render_query_human(r, "q")
+    assert "capacity:" not in out
+
+
+def test_render_capacity_thousand_separators():
+    """Large prompts must format with thousand-separator commas so
+    a 60K prompt reads as '60,000' not '60000' — operator legibility
+    on the daily render."""
+    r = _result(
+        prompt_chars={
+            "system_prompt": 1500,
+            "grounding_reminder": 0,
+            "user_question": 50,
+            "evidence_or_context": 60000,
+            "messages_total": 61550,
+        },
+        answer_chars=1234,
+    )
+    out = _render_query_human(r, "q")
+    assert "61,550 chars" in out
+    assert "60,000" in out
+    assert "1,234 chars" in out