zfs: fix panic due to tqid overflow

[mjg]

Summary

The 32-bit counter eventually wraps to 0 which is a sentinel for invalid id.

Make it 64-bit on LP64 platforms and 0-check otherwise.

Note: Linux counterpart uses id stored per queue instead of a global.
I did not check going that way is feasible with the goal being the minimal fix doing the job.

Once this is sorted out I'll post a review on github.

[mav]

Makes sense to me. I just hope that situation of ID duplicate due to wraparound it too unlikely on 32-bit even considering taskq_dispatch_delay(). Though not sure who really used any 32bit platforms recently, even considering recent i386 KVA increase to 4GB.