daemon/worker: fix a memory leak

Discovered case: TCP write towards upstream fails due to ECONNRESET, and on this place of code we "forget" the whole qr_task and thus its corresponding kr_request, so it remains unanswered and using memory.

added bug label

I have more improvements in progress, but this minimal part seemed important enough to MR separately. (though the verbose-log parts could also be omitted)

With this code version I can no longer reproduce the leaks or any other sanitizer problems; heavily tested on https://gitlab.nic.cz/knot/knot-resolver/-/jobs/553090 (the verbose-log part differed there)

approved this merge request

added 19 commits

• 8fe81a11...c6ab0793 - 18 commits from branch master
• 5b2ba2d2 - daemon/worker: fix a memory leak

Compare with previous version

mentioned in commit c32dee26

merged

This change is most likely what triggers an occasional corruption among our uv_timer_t, probably some of them may destruct incorrectly which leads to later crashes when libuv operates on the heap of timers. (As the crashes appear to be out of context, coredumps probably won't help that much to debug this.)

I managed to reproduce that corruption even before this MR (on v5.3.2), but it's still possible that this fix makes it more likely somehow.