attempt to ensure metacontroller finalizers always run

[ikreymer]

Looking into this issue more, it seems the ttlSecondsAfterFinished set to 0 is likely the main culprit for finalizers being skipped for these resources, as the job is cleaned up right away. However, to guarantee the finalizers are always run, they should also declared as part of the object metadata at creation time for the resources that have these hooks.

This PR:

• set ttlSecondsAfterFinished to 10 for replica jobs, 1 for other background jobs (to avoid breaking tests which wait for these jobs to be finalized)
• explicitly declares the appropriate metacontroller finalizer hook on each resource that has a finalize hook (crawljob, collindex, background-job) at resource creation time

Should fix #3365

[ikreymer]

Set the ttl lower for replica and other background jobs, as some code paths (like collection update) does wait for the jobs to finish, in tests. Perhaps keeping ttl at 0 or 1 does make sense, as long as the finalizer is set on create, as the whole point is to respond quickly as soon as the job is finished

[tw4l]

I'm not convinced ttlSecondsAfterFinished is playing a part here (though I'm also fine with the small changes to those values here, just haven't seen evidence that plays into the bug), but explicitly setting the finalizers seems like a good step.

One suggestion where I think we need another finalizer added, and will need to do the same in #3577.

[ikreymer]

I'm not convinced ttlSecondsAfterFinished is playing a part here (though I'm also fine with the small changes to those values here, just haven't seen evidence that plays into the bug), but explicitly setting the finalizers seems like a good step.

I think the finalizers are the real fix, so maybe no need to change the ttl, though it probably makes it more likely to happen. Removed the ttl changes for now so we can check if the finalizers definitively fix this.