nvme: fix block count and data size logic

nvme: fix block count and data size logic

The required block count and data size is not automatically set
for a namespace device in submit_io(). Also it is the user specified
block count and not the actual required block count that is passed
to the ioctl data structure, often leading to nvme passthrough io
errors. For e.g. an nvme read on a 512b block size namespace device
with data size set to 4K ends up with the below error:

nvme read /dev/nvme0n1 -s 0 -z 4096
NVMe status: Data SGL Length Invalid: The length of a Data SGL is too
short or too long and the controller does not support SGL transfers
longer than the amount of data to be transferred(0x400f)

This read is successful only when the appropriate block count is
passed from the command line:

nvme read /dev/nvme0n1 -s 0 -z 4096 -c 7
read: Success

So fix this by deriving the required block count and data size based
on the logical block size and passing them appropriately to the
respective ioctl data structure. And while we are at it, remove a
couple of superfluous braces in this part of the code.

Signed-off-by: Martin George <marting@netapp.com>