mds: count purge queue items left in journal
MDS purge queue didn't have a perf counter to record how many items still left in journal. Even when MDS restarted, there was no any hint to know how many inodes haven't been really deleted from disks, so it is not easy to monitor disk usage, do flow control, etc.
Now we add a new perf counter called "l_pq_item_in_journal". Although it is an approximation value, I think we can tolerate with this and also it doesn't need to be precise.