Fix reload 100% CPU spinning after failed reload

[wtoorop]

After a failed reload, some UDB pointers were cleared that were still linked in the UDB. This fix unlinks them in all the correct places.

This may be the fix to issue #417

[wcawijngaards]

It makes sense that on failure, the allocated objects need to be taken out of use. Hopefully that fixes the task file corruption issue.

[wtoorop]

It makes sense that on failure, the allocated objects need to be taken out of use. Hopefully that fixes the task file corruption issue.

@wcawijngaards I want to do it differently anyway.

Although I do think this will fix the 100% spinning CPU at the load process (also because the last_task is not used by the forked load process if one of the process xfr tasks fails), I do still feel uncomfortable that the udb_ptr's are implicitly copied by the fork. I would rather have them unlinked after the first iteration of processing the non-transfer tasks, so that the failure cases don't need to to touch them at all.

I'd like to try another iteration where both last_task and xfrs2process are conveyed to the load process after fork on the task list itself. Hopefully I can work on that this afternoon and we can do an emergency nsd-4.11.1 quickfix release early next week (when results from @pattai confirm this fixed the issue).

[wtoorop]

Closing this in favor of the simpler approach in #424