nvme: use the atomic queue limits update API

Changes the callchains that update queue_limits to build an on-stack
queue_limits and update it atomically.  Note that for now only the
admin queue actually passes it to the queue allocation function.
Doing the same for the gendisks used for the namespaces will require
a little more work.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Keith Busch <kbusch@kernel.org>