test(ext/ffi): Increase timeout value in event loop integration test callback

[aapoalas]

The test has flakes for essentially the same reason as what it is trying to test: How asynchronous work started from FFI callbacks gets resolved.

Due to Deno using the v8::MicrotasksPolicy::kAuto for handling microtasks, the microtask queue is drained synchronously when the script call depth decrements to zero. FFI callbacks coming from foreign threads get called one at a time from an empty scope created expressly for the purpose of calling that particular callback. As a result, each FFI callback immediately drains the microtask queue upon its completion.

The test uses two calls, queueMicrotask and setTimeout to respectively push one microtask into the queue and push one macrotask into their separate queue. However, setTimeout isn't exactly guaranteed to create a macrotask. A short enough timeout can cause the task to become a microtask instead. The 10 ms timeout had been chosen as a safe bound but apparently it is not quite safe enough. The flake occurs when the setTimeout manages to inch itself into the microtask queue:

The test expects that first the "Sync" will be logged from the callback directly, then "Async" from the queueMicrotask microtask. Then, the Rust code calling the callback will log its "STORED_FUNCTION called", after which finally the setTimeout task logs "Timeout". When setTimeout gets into the microtask queue, it instead logs before the Rust code is returned to.

From a JS point of view this would look like this to:

(() => {
  console.log("Sync");
  queueMicrotask(() => console.log("Async"));
  setTimeout(() => console.log("Timeout"), 10);
})();
console.log("Done");

[dsherret]

LGTM. Thanks, @aapoalas!