Make the regression test for pg_eucjp stable.

Previously, the regression test for the pg_eucjp function output
the EUC_JP characters in the order of the character itself.
However, this order could vary depending on the platform,
causing the test to fail unpredictably.

To address this issue, this commit updates the test so that
it outputs the characters in the order of each character's code instead.
This ensures that the output is consistent across platforms and
guarantees that the test is stable.

No functional changes are made to the pg_eucjp function.

expected/pg_eucjp.out

@@ -88,7 +88,7 @@ SELECT count(*) FROM all_eucjp
      0
 (1 row)
 
-SELECT * FROM all_eucjp ORDER BY eucjp;
+SELECT * FROM all_eucjp ORDER BY code1, code2, code3;
  code1 | code2 | code3 |  eucjp   
 -------+-------+-------+----------
  x00   | x00   | x00   | 

sql/pg_eucjp.sql

@@ -32,7 +32,7 @@ CREATE TABLE all_eucjp(code1 text, code2 text, code3 text, eucjp text);
 INSERT INTO all_eucjp SELECT * FROM pg_all_eucjp();
 SELECT count(*) FROM all_eucjp
     WHERE pg_eucjp(code1::bit(8), code2::bit(8), code3::bit(8)) <> eucjp;
-SELECT * FROM all_eucjp ORDER BY eucjp;
+SELECT * FROM all_eucjp ORDER BY code1, code2, code3;
 
 SELECT pg_eucjp('xb6', 'xe5') || pg_eucjp('xfc', 'xf9') || pg_eucjp('xad', 'xa9') || pg_eucjp('xfa', 'xdb');