fix(model-resolution): normalize model format and remove dead config flag

src/shared/model-resolution-pipeline.test.ts

@@ -0,0 +1,25 @@
+import { describe, expect, test } from "bun:test"
+import { resolveModelPipeline } from "./model-resolution-pipeline"
+
+describe("resolveModelPipeline", () => {
+  test("does not return unused explicit user config metadata in override result", () => {
+    // given
+    const result = resolveModelPipeline({
+      intent: {
+        userModel: "openai/gpt-5.3-codex",
+      },
+      constraints: {
+        availableModels: new Set<string>(),
+      },
+    })
+
+    // when
+    const hasExplicitUserConfigField = result
+      ? Object.prototype.hasOwnProperty.call(result, "explicitUserConfig")
+      : false
+
+    // then
+    expect(result).toEqual({ model: "openai/gpt-5.3-codex", provenance: "override" })
+    expect(hasExplicitUserConfigField).toBe(false)
+  })
+})