gemini: minimal thinking on correlation, token accounting per call

A day of replay runs (server-26#170) spent ~$5 of Gemini on ~7 two-hour
windows (~$0.70 per 290 calls) — several dollars a day per live deployment
for correlation alone — and nothing could say where it went (#45). Gemini
3.x thinks by default and bills it as output; the deprecated
google-generativeai SDK these calls used cannot set a thinking level.

- app/internal/gemini.py: every Gemini call (correlation + transcript
  correction) goes through google-genai with JSON mode, an explicit
  thinking level, and logs in/out/thinking tokens. A model that rejects
  the level is retried without it once and remembered, so the tier is
  never lost to a config param. API failures still raise for ai_health.
- correlator: thinking_level "minimal" (a link/new/orphan choice).
  transcript correction: "low" until a replay shows minimal is safe.
- replay: runs record real Gemini token usage (metrics.gemini_usage),
  shown in the Replay tab.
- requirements: google-genai.

c2-core: 474 pass. Frontend typecheck not run (no Node on this box).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

drb-frontend/components/admin/ReplayTab.tsx

@@ -361,6 +361,14 @@ function RunDetail({ run }: { run: ReplayRun }) {
           paths: {Object.entries(m.corr_path).map(([k, v]) => `${k} ${v}`).join(" · ")}
         </p>
       )}
+      {m?.gemini_usage && Object.keys(m.gemini_usage).length > 0 && (
+        <p className="text-xs font-mono text-gray-500">
+          Gemini tokens:{" "}
+          {Object.entries(m.gemini_usage)
+            .map(([k, u]) => `${k} ${u.calls} calls, in ${u.in.toLocaleString()} / out ${u.out.toLocaleString()} / thinking ${u.thinking.toLocaleString()}`)
+            .join(" · ")}
+        </p>
+      )}
       {m?.ai_failures && Object.keys(m.ai_failures).length > 0 && (
         <p className="text-xs font-mono text-amber-400">
           AI failures: {Object.entries(m.ai_failures).map(([k, v]) => `${k} ×${v}`).join(" · ")}