Add closed state to Mailbox

[ntkme]

This is an alternative solution to the same problem described in #25 for sass/dart-sass#2279

On high level we have following logic in dart-sass, and we need a way to reliable stop the loop via mailbox:

Main Isolate:

isolate.kill();
try {
  mailbox.put(Uint8List(0));
} on StateError catch (_) {
  // When delivery of empty message fails due to existing message, the shutdown might lockup.
}

Worker Isolate:

do {
  Uint8List data = mailbox.take();
  if (data.isEmpty) {
    break;
  }
  // do work with data
} while (true);

First, Isolate.kill(priority: Isolate.immediate) won't work when mailbox.take() is blocking. So, instead we call Isolate.kill() first, and then send an empty message to signal the stopping. However, because the sending of empty message can fail, and then depending on the event loop timing, the worker isolate may process the isolate kill event before entering the loop one more time, which works as expected (killed before next iteration of loop); or it may get to the next iteration too quickly before the isolate kill event is processed, and thus cause the VM to lock up because the next mailbox.take() blocks the processing of kill event, causing the isolate to be appear non-responsive.

---

Adding close method as designed in this PR will allow us to do:

Main Isolate:

isolate.kill();
mailbox.close();

Worker Isolate:

do {
  Uint8List data;
  try {
    data = mailbox.take();
  } on StateError catch (_) {
    break;
  }
  // do work with data
} while (true);

Because mailbox.close() is guaranteed to deliver the signal via StateError on the ongoing or next mailbox.take(), we don't have the problem of failing to send empty message.

[ntkme]

@devoncarew Is this something you can help review, or help identify a reviewer, as it appears @mraleph might not be available.

[devoncarew]

@devoncarew Is this something you can help review, or help identify a reviewer, as it appears @mraleph might not be available.

Unfortunately I'm not familiar with this package; @mraleph will be back in a week (@mkustermann - the other likely reviewer - is OOO at the moment as well).

[ntkme]

@mraleph I've updated the documentation. Can you please take another look when you get a chance?