Write tests for fetchUpstream

PR #214 was merged. fetchUpstream now returns null when a request times out instead of hanging — a real improvement. But there are zero tests covering it.

That means the next person to touch this function has no safety net. A refactor, a well-intentioned cleanup, a merge that goes slightly wrong — any of these could silently undo the timeout behavior with no warning from CI.

Your job is to write the tests that make this behavior explicit and permanent.

The situation

request.ts is already implemented and working. request.test.ts exists but is empty. The function has three behaviors worth pinning down:

• It returns a Response when the upstream answers normally.
• It returns null — not throws — when the request times out.
• It always clears the internal timer after the call, whether it succeeded or not.

The third one is subtle. A timer that isn't cleared keeps the Node.js process alive longer than it should. It's the kind of thing that causes flaky tests in other files.

What you'll do

1. Read request.ts to understand what the function does and when it returns null.
2. Write three tests in request.test.ts covering the behaviors above.
3. Run the tests and make sure all three pass.

Done when

• The happy path test passes: a normal response is returned correctly.
• The timeout test passes: the function returns null and does not throw.
• The cleanup test passes: clearTimeout is called exactly once after every call.

Mock fetch globally — don't make real network requests. For the timeout case, reject with a DOMException named 'AbortError', which is what AbortController raises when it aborts a signal.