btrfs: add fast path for single device io in __btrfs_map_block

There is no need for most of the btrfs_io_context when doing I/O to a
single device.  To support such I/O without the extra btrfs_io_context
allocation, turn the mirror_num argument into a pointer so that it can
be used to output the selected mirror number, and add an optional
argument that points to a btrfs_io_stripe structure, which will be
filled with a single extent if provided by the caller.

In that case the btrfs_io_context allocation can be skipped as all
information for the single device I/O is provided in the mirror_num
argument and the on-stack btrfs_io_stripe.  A caller that makes use of
this new argument will be added in the next commit.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David Sterba <dsterba@suse.com>