Provide jobs with an AbortSignal to attempt to gracefully interrupt jobs after shutdown timeout occurs

[jbielick]

Purpose

A job may be interrupted when a worker shuts down. In this case there are two mechanisms to ensure graceful interruption: the shutdown timeout and the execution context AbortSignal. The shutdown timeout is configured in WorkerOptions.timeout. When a worker is instructed to stop (via process signal or server message), it will stop accepting new work (e.g. quiet) and wait the configured duration for any in-progress jobs to complete uninterrupted. If this duration elapses and jobs are still in progress, these jobs will receive an AbortSignal via Context.signal. All jobs will be FAILed on the Faktory server, allowing them to retry later.

The abort signal can be used to interrupt asynchronous processes and perform some cleanup tasks before an abrupt exit (process.exit). After the abort signal is sent, a job will have 3 seconds to perform cleanup before the process is abruptly exited.

Approach

• Private AbortController created upon .work()
• After stop() and shutdown timeout, in-progress jobs are sent an abort signal and subsequently FAILed on the server.
• 3 seconds delay is added after abort signal is sent to allow cleanup
• exit code when jobs are aborted at end of grace period is 1

Example - A long-running subprocess:

faktory.register("JobWithAbort", (...args) => async ({ signal }) => {
  try {
    await execa("ffmpeg", [/* arg1, arg2, ..., argN */], { cancelSignal: signal });
  } catch (e) {
    if (e.code === "ABORT_ERR") {
      // Remove some tempfiles or other type of cleanup...
      // Propagating the ABORT_ERR is not necessary, the job will be FAILed if it was in-progress 
      // at the end of the shutdown timeout
    }
  }
});

Closes #409
Closes #251