feat(coach): resolve the turn's thinking depth from the person and the operator

The chat route hands the turn the level resolveCoachTurnReasoningLevel
returns: the stored Coach preference, absent read as medium, after the
operator's switch and cap. The turn still applies the operator-paid
ceiling once its chain is known. The controls loader no longer throws
when the request cache cannot serve it; it fails closed to off. The
admin guard pins the route to the resolver.

src/app/api/insights/chat/__tests__/golden-harness.ts

@@ -175,6 +175,8 @@ export function resetGolden(): void {
     },
   );
   m.getEvent.mockReturnValue({
+    // A request event, so the per-request cache memoises as in production.
+    getKind: () => "request",
     setError: (err: Error) =>
       record("event.setError", { message: err.message }),
   });

src/app/api/insights/chat/__tests__/golden-transcripts.test.ts

@@ -63,6 +63,8 @@ vi.mock("@/lib/db", () => ({
     user: { findUnique: h.m.userFindUnique, update: h.m.userUpdate },
     coachConversation: { findFirst: h.m.conversationFindFirst },
     coachMessage: { findMany: h.m.coachMessageFindMany },
+    // v1.41 — the operator's reasoning controls; an untouched instance.
+    appSettings: { findUnique: vi.fn(async () => null) },
   },
 }));
 vi.mock("@/lib/rate-limit", () => ({

src/app/api/insights/chat/route.ts

@@ -37,10 +37,8 @@ import { coachChatRequestSchema } from "@/lib/ai/coach/types";
 import { listConversations } from "@/lib/ai/coach/persistence";
 import { detectRefusal } from "@/lib/ai/coach/refusal";
 import { runCoachTurn } from "@/lib/ai/coach/turn/pipeline";
-import {
-  DEFAULT_REASONING_LEVEL,
-  type ReasoningLevel,
-} from "@/lib/ai/reasoning/levels";
+import type { ReasoningLevel } from "@/lib/ai/reasoning/levels";
+import { resolveCoachTurnReasoningLevel } from "@/lib/ai/reasoning/controls";
 import { streamProviderError, streamRefusal } from "@/lib/ai/coach/turn/sse";
 
 /**
@@ -182,10 +180,13 @@ async function handleChatRequest(request: NextRequest): Promise<Response> {
   }
 
   // v1.41 — the person's effective reasoning level: their choice after the
-  // operator's switch and highest level. Integration points this one call
-  // site at the shared resolver (`src/lib/ai/reasoning/resolve.ts`); until
-  // then every turn thinks at the default level.
-  const reasoningLevel: ReasoningLevel = DEFAULT_REASONING_LEVEL;
+  // operator's switch and highest level, from the shared resolver. Who pays
+  // is not known until the turn resolves its chain, so the turn applies the
+  // operator-paid ceiling itself (`turn/reasoning.ts`); the provider's own
+  // ability is left to its client.
+  const reasoningLevel: ReasoningLevel = await resolveCoachTurnReasoningLevel(
+    auth.user.coachPrefsJson,
+  );
 
   // v1.41 — the person's turn slot. The counter is the rate-limit row's own
   // (an atomic upsert), given back when the turn is over; a refused slot is