Measure the counterpart honestly: warm models, vectorised search, scope split

Three defects made the external baseline numbers and labels wrong.

- bench/systems.py: _sentence_model rebuilt the model on every query, so dense arm
  latency was mostly model loading and the latency-vs-size curve was flat because a
  constant dominated it. Models are now cached per process.
- bench/counterpart.py: visible() and search() walked 404k Python objects per query.
  Per-item holder/source/time columns are built once and the query path is numpy;
  12.5 queries/s instead of 0.7.
- bench/metrics.py: the aggregate key was named wrong_holder_confidence_unanswerable
  while it counts top hits that ARE from the asking holder. BASELINE.md read that
  backwards. Renamed to top_hit_from_holder_unanswerable and the sentence rewritten.
- bench/run_external.py: the by-modality table read row['answerable'] but aggregate
  emits n_answerable, so the writer died after scoring finished. Each store is now
  scored twice, [holder scope] and [whole corpus], so the gain from the holder
  boundary is separable from the gain from ranking; progress lines every 500 queries.

Verification: pytest bench/tests -q -> 61 passed; ruff check bench/ -> all checks
passed; mypy (fresh cache) -> no issues in 25 source files.

bench/metrics.py

@@ -142,7 +142,7 @@ def aggregate(scores: Sequence[QueryScore]) -> dict[str, float | int]:
         "abstention_rate_unanswerable": (
             sum(1 for s in unanswerable if s.abstained) / len(unanswerable) if unanswerable else 0.0
         ),
-        "wrong_holder_confidence_unanswerable": (
+        "top_hit_from_holder_unanswerable": (
             sum(1 for s in holder_known if s.top_hit_from_holder) / len(holder_known)
             if holder_known
             else 0.0