[14.x] Job expiration based on job timeout, remove the need for retry_after

[SjorsO]

Last attempt at trying to get #61158 merged, this time targeting the next major release.

I'm aware this is a risky PR. It modifies core queue behavior, which is rock solid apart from the retry_after quirks. Still, I think the juice is worth the squeeze. It has worked for everything I've tested, it's backwards compatible, it removes a footgun in queued jobs, and it removes the need to explain the confusing retry_after concept in the docs.

This PR is mostly beneficial for projects that have increased their retry_after, which in my experience is common. I've worked on multiple projects where retry_after was set to 1 hour.