fix(core): fix another scenario that causes hanging (11.x)

[eliangcs]

Related to #509, #490, and https://github.com/zapier/zapier/issues/53430, this addresses another scenario that causes the Lambda handler to hang/freeze.

Suppose a converted app has this (bad) legacy script:

var Zap = {
  bad_post_write: function(bundle, callback) {
    z.request({ ... }, function(err, response) {
      if (err) {
        callback(err);
      } else {
        callback(null, z.JSON.parse(response.content));
      }
    });
    callback(null, { message: 'ok' }); // double callback?!
  }
};

Notice how the post_write function returns immediately by calling callback(null, { message: 'ok' }) while makes an outbound z.request() and calls callback asynchronously.

When legacy-scripting-runner promisifies this post_write function, the callback in the z.request will come after the Lambda handler returns. Since we only end the logger once when the Lambda handler finishes, the HTTP logger corresponding to the z.request call, which comes after the Lambda handler, will stay open and never end until a long timeout.

To fix, we need to make sure to end any loggers allocated after the Lambda handler returns.

[eliangcs]

Fixed a problem while I was here. We should call either resolve or reject but not both.

[eliangcs]

This code has essentially the same effect as the "double callback" example I wrote in the PR description.

[eliangcs]

@xavdid sorry I haven't reviewed your PRs #512 and #511. I wanted to merge this PR to 11.x before your PRs targeting 12.x. I'll make sure to review your PRs my tomorrow! 🙇

[eliangcs]

@xavdid can you also review/approve this PR when you're online? Thanks!

[xavdid]

Ah, sorry! I approved the other one and I guess I missed this one in the shuffle.

I'll hold off merging mine so you have a clean path to get these out.