Once we resume a query, we should clear DNS_FETCHOPT_TRYSTALE_ONTIMEOUT
from the options to prevent triggering the stale-answer-client-timeout
on subsequent fetches.
If we don't this may cause a crash when for example when prefetch is
triggered after a query restart.
(cherry picked from commit c0dc5937c7)
a2ec3a1e4c