btrfs-progs: replace: fix an unexpected new line when replace failed

[BUG]
When a device replace failed, e.g. try to replace a device on a RO
mounted btrfs, the error message is incorrectly broken into two lines:

 [adam@btrfs-vm ~]$ sudo btrfs replace start -fB 1 /dev/test/scratch3  /mnt/btrfs/
 Performing full device TRIM /dev/mapper/test-scratch3 (10.00GiB) ...
 ERROR: ioctl(DEV_REPLACE_START) failed on "/mnt/btrfs/": Read-only file system

 [adam@btrfs-vm ~]$

Note the newline after the "Read-only file system" error message.

[CAUSE]
Inside cmd_replace_start(), if the ioctl failed we need to handle the
error messages different depeneding on start_args.result.

If the result is not BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_RESULT we will
append extra info to the error message.

But the initial error message is using error(), which implies a newline.

This results the above incorrectly splitted error message.

[FIX]
Instead of manually appending an extra reason to the existing error
message, just do different output depending on the start_args.result in
the first place.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>

Author: adam900710

cmds/replace.c

@@ -319,12 +319,11 @@ static int cmd_replace_start(const struct cmd_struct *cmd,
 	ret = ioctl(fdmnt, BTRFS_IOC_DEV_REPLACE, &start_args);
 	if (do_not_background) {
 		if (ret < 0) {
-			error("ioctl(DEV_REPLACE_START) failed on \"%s\": %m", path);
-			if (start_args.result != BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_RESULT)
-				pr_stderr(LOG_DEFAULT, ", %s\n",
-					replace_dev_result2string(start_args.result));
+			if (start_args.result == BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_RESULT)
+				error("ioctl(DEV_REPLACE_START) failed on \"%s\": %m", path);
 			else
-				pr_stderr(LOG_DEFAULT, "\n");
+				error("ioctl(DEV_REPLACE_START) failed on \"%s\": %m, %s",
+				      path, replace_dev_result2string(start_args.result));
 
 			if (errno == EOPNOTSUPP)
 				warning("device replace of RAID5/6 not supported with this kernel");