fix(reasoning): preserve effort through Codex translations

Carry Claude reasoning_effort/reasoning into OpenAI Chat, map into
OpenAI Responses reasoning.effort, and keep request-level effort
(incl. xhigh) across tool-result turns instead of collapsing to high.

Co-authored-by: Cursor <cursoragent@cursor.com>

tests/translator/thinking-unified.test.js

@@ -123,6 +123,10 @@ describe("applyThinking per provider format", () => {
     const out = apply("openai", "gpt-5(low)", { reasoning_effort: "high" }, "openai");
     expect(out.reasoning_effort).toBe("low");
   });
+  it("openai keeps xhigh for reasoning models", () => {
+    const out = apply("openai", "gpt-5.3-codex", { reasoning_effort: "xhigh" }, "codex");
+    expect(out.reasoning_effort).toBe("xhigh");
+  });
 });
 
 describe("extractReasoningText (response shapes)", () => {

tests/unit/provider-thinking-config.test.js

@@ -0,0 +1,17 @@
+import { describe, it, expect } from "vitest";
+import { normalizeThinkingConfig } from "../../open-sse/services/provider.js";
+
+describe("normalizeThinkingConfig", () => {
+  it("keeps openai reasoning_effort on non-user turns", () => {
+    const body = {
+      messages: [{ role: "assistant", content: "ok" }],
+      reasoning_effort: "xhigh",
+      thinking: { type: "enabled" },
+    };
+
+    normalizeThinkingConfig(body);
+
+    expect(body.reasoning_effort).toBe("xhigh");
+    expect(body.thinking).toBeUndefined();
+  });
+});