patch: Indent error output when sort test fails.

[Petter Reinholdtsen]

[Ulrich Drepper]
> You still haven't provided a reason.  Why should this output format
> be better?

Ah, I thought the "to make it easier to visually group the output"
reason was sufficient.

It is harder to read and immediately know how to group the current
output, without any visual grouping, then it is to read a version
where the text belonging to the failing locale is indented.

This version is harder to read:

de_DE.ISO-8859-1 OK
en_US.ISO-8859-1 OK
nn_NO.ISO-8859-1 OK
nb_NO.ISO-8859-1 FAIL
--- nb_NO.in  2003-04-30 07:03:35.000000000 +0000
+++ /usr/src/glibccvs/libc/BUILD/localedata/nb_NO.xout        2003-05-29 15:06:36.000000000 +0000
@@ -24,6 +24,6 @@
 x
 y
 z
+<E5>
 <E6>
 <F8>
-<E5>
da_DK.ISO-8859-1 OK
hr_HR.ISO-8859-2 OK
sv_SE.ISO-8859-1 OK
no_NO.ISO-8859-1 OK

This version is easier to read:

de_DE.ISO-8859-1 OK
en_US.ISO-8859-1 OK
nn_NO.ISO-8859-1 OK
nb_NO.ISO-8859-1 FAIL
  --- nb_NO.in  2003-04-30 07:03:35.000000000 +0000
  +++ /usr/src/glibccvs/libc/BUILD/localedata/nb_NO.xout        2003-05-29 15:06:36.000000000 +0000
  @@ -24,6 +24,6 @@
   x
   y
   z
  +<E5>
   <E6>
   <F8>
  -<E5>
da_DK.ISO-8859-1 OK
hr_HR.ISO-8859-2 OK
sv_SE.ISO-8859-1 OK
no_NO.ISO-8859-1 OK