perf(hunter): cacheable message order; adaptive step budget in the sink verifier

Token efficiency, measured on 180 verified pgAdmin candidates: stage 2 costs
$0.0139 per candidate, 90% of the spend, because the conversation is re-sent on
every turn and grows with each tool result.

- Context files now precede the question in the hunter's first message. DeepSeek
  bills a cached prefix at 1/30 of a fresh one, by prefix; with the question first,
  every candidate asked about the same file (about 5 per file) broke the prefix on
  its first line and paid for the file again. A test pins the order.
- The verifier runs each candidate at 3 steps and repeats it once at 6 only when
  the hunt spent every step on tool calls without a verdict. On the seven declared
  v2 sites, 3 steps found 5 and 8 steps found 6; most candidates conclude in 3.
- Per-candidate token usage (prompt, cached, completion, calls) and escalation
  are recorded, and each case reports its cache share, so the next efficiency
  decision is made on numbers.

Validation on the 180 pgAdmin candidates plus the 22 known sites (about $1)
waits for account credit: DeepSeek returns 402 at the moment.

Co-Authored-By: Claude Fable 5.1 <[email protected]>
Claude-Session: https://claude.ai/code/session_01EmM1j46qgKMh3zqD1QMwEo

tests/test_tool_hunter.py

@@ -365,3 +365,14 @@ def test_the_final_answer_is_asked_plainly_and_a_wrapped_array_is_read(tmp_path)
     found = tool_hunter.run_tool_hunter(tmp_path, [spot], client=Client(), model="m", max_steps=3)
     assert [(f.path, f.line) for f in found] == [("app.py", 2)]
     assert seen[-1] == {"tools": [], "json_object": False}, seen[-1]
+
+
+def test_context_files_precede_the_question_so_the_prefix_caches(tmp_path) -> None:
+    """DeepSeek bills a cached prefix at 1/30 of a fresh one, by prefix. With the question first, every candidate
+    asked about the same file broke the prefix on its first line and paid for the file again."""
+    from openultrasast import tool_hunter
+
+    (tmp_path / "app.py").write_text("def run(q):\n    return eval(q)\n")
+    message = tool_hunter._user_message(tmp_path, [], "Judge the eval in run.", ["app.py"], 4000)
+    assert message.index("--- app.py") < message.index("def run(q)") < message.index("Judge the eval in run.")
+    assert message.endswith("Judge the eval in run.")