btrfs: do not utilize goto to implement delayed inode ref deletion

[PROBLEM]
The function __btrfs_update_delayed_inode() is doing something not
meeting the code standard of today:

	path->slots[0]++
	if (path->slots[0] >= btrfs_header_nritems(leaf))
		goto search;
again:
	if (!is_the_target_inode_ref())
		goto out;
	ret = btrfs_delete_item();
	/* Some cleanup. */
	return ret;

search:
	ret = search_for_the_last_inode_ref();
	goto again;

With the tag named "again", it's pretty common to think it's a loop, but
the truth is, we only need to do the search once, to locate the last
(also the first, since there should only be one INODE_REF or
INODE_EXTREF now) ref of the inode.

[FIX]
Instead of the weird jumps, just do them in a stream-lined fashion.
This removes those weird labels, and add extra comments on why we can do
the different searches.

Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>