fix: Long retries

[nzakas]

ESLint reported that some users were still getting EMFILE errors even with using Retry.

The primary problem is that #processQueue was being called recursively, and when a lot of retries were necessary, this caused a stack overflow error.

There was another problem where a failed task was pushed to the front of the queue, allowing it to fail repeatedly while other tasks in the queue sat without being evaluated. Those tasks would then end up timing out and throwing the EMFILE error that they held.

This PR fixes both issues:

1. Ensures that #processQueue is always called either with setTimeout() or in a promise resolver
2. Failed tasks are pushed to the back of the queue to wait for another turn rather than the front.

I also added a test for fails for long retries prior to this fix. This will now be run as part of CI.

Refs eslint/eslint#18977