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-c2-core/app/internal/replay.py

@@ -469,6 +469,9 @@ async def _run(run_id: str, org_id: str, calls: list[dict], mode: str,
     fl_token = force_flags(_flags_for(mode))
     ai_failures: list = []
     ai_token = ai_health.collect_sandbox_failures(ai_failures)
+    from app.internal import gemini
+    usage: dict = {}
+    usage_token = gemini.collect_usage(usage)
     try:
         sem = asyncio.Semaphore(PREFETCH)
 
@@ -561,6 +564,7 @@ async def _run(run_id: str, org_id: str, calls: list[dict], mode: str,
         metrics = compute_metrics(incidents, sb_calls)
         metrics["est_cost_usd"] = _running_cost(progress, metrics, mode)
         metrics["ai_failures"] = dict(Counter(f"{f['tier']}: {f['problem']}" for f in ai_failures))
+        metrics["gemini_usage"] = usage
     except Exception as e:
         status = "failed"
         errors.append(f"run: {type(e).__name__}: {e}"[:300])
@@ -568,6 +572,7 @@ async def _run(run_id: str, org_id: str, calls: list[dict], mode: str,
         logger.error(f"Replay {run_id} failed: {e}")
     finally:
         ai_health._sandbox_failures.reset(ai_token)
+        gemini.reset_usage(usage_token)
         unforce_flags(fl_token)
         fstore.exit_sandbox(sb_token)
         _cancel.discard(run_id)