Previously, taskmgr was using ISC_LIST for keeping the enqueued tasks
and the ISC_LIST have an inherent property that the task would not get
enqueued twice. When taskmgr switched to using netmgr, we are using
isc_queue_t which can enqueue same item multiple times, and this was
causing problem when the previous isc_task_run() would find the task to
be empty and shutting down. It would directly destroy the task and
leave the following netievents with already freed task.
This commit adds references counting around task_ready(++) and
task_run(--) and adds isc_task_ready() public API wrapper around
task_ready, so we can call it from netmgr when the event quantum would
be reached.
f6289afc97