diigo-cache: pace diigo calls only; cache hits don't wait

page() ran the 500 ms pacing gate before the cache check, so every hit
waited too: a fill rerun skipping 166 cached pages took ~80 s and
looked like a start from scratch. The gate now sits after the hit
check (re-checking storage once through it, in case a request ahead
had just etched the page).

Regression test: the fixed-clock test cache records its sleeps; a miss
sleeps once, three hits add none (failed before: 4 sleeps).

Generated-by: Claude Opus 5.5

diigo-cache/tests/caches.test.ts

@@ -12,6 +12,7 @@ const WEBAPP_ID = "fake-webapp-id";
 type TestEnv = {
 	CACHE_DO: DurableObjectNamespace<import("../src/index.js").DiigoCache>;
 	PUBLIC_DO: DurableObjectNamespace<import("../src/index.js").PublicCache>;
+	CLOCKED_DO: DurableObjectNamespace<import("./test-worker.js").FixedClockCache>;
 	WEBAPP: { calls(): Promise<number> };
 };
 const testEnv = env as unknown as TestEnv;
@@ -109,6 +110,18 @@ describe("page size", () => {
 	});
 });
 
+describe("pacing", () => {
+	it("paces diigo calls only: a cache hit doesn't wait", async () => {
+		const stub = testEnv.CLOCKED_DO.get(testEnv.CLOCKED_DO.newUniqueId());
+		await stub.setup({ webappId: WEBAPP_ID });
+		await stub.page(1, PAGE_SIZE);
+		const afterMiss = (await stub.sleeps()).length;
+		expect(afterMiss).toBeGreaterThan(0);
+		for (let i = 0; i < 3; i += 1) expect((await stub.page(1, PAGE_SIZE)).cached).toBe(true);
+		expect(await stub.sleeps()).toHaveLength(afterMiss);
+	});
+});
+
 describe("eviction", () => {
 	it("a cache miss leaves no undisposed RPC stubs to block eviction", async () => {
 		const stub = testEnv.CACHE_DO.get(testEnv.CACHE_DO.newUniqueId());