LoadJob::onNewEntry() receives an Entry pointer. When we kill a LoadJob, the
libarchive plugin deletes all the emitted entries, so onNewEntry() (which is
queued in the events) gets a pointer that points to already free'ed memory.
This patch just delays the deletion of entries by using deleteLater().
This ensures that all the slots are executed first.
An alternative fix could be using a DirectConnection for the entry ->
onNewEntry connection...