Fix instability with WAL fsync test in stats.sql

A backend using wal_sync_method set to "open_sync" or "open_datasync"
would fail the test checking the WAL sync data in pg_stat_io.  These
modes guarantee that a sync is done when WAL is written to disk, and the
data checked by the test is not incremented in this case,
issue_xlog_fsync() doing nothing.

Oversight in commit a051e71e28.

Author: Sami Imseih <samimseih@gmail.com>
Discussion: https://postgr.es/m/CAA5RZ0uxwg3xAi4nvdBMJ-zJQEeyg+RotuU+ebM2F6CKmnvaYA@mail.gmail.com

src/test/regress/expected/stats.out

@@ -1456,6 +1456,7 @@ SELECT :io_sum_wal_normal_after_writes > :io_sum_wal_normal_before_writes;
 (1 row)
 
 SELECT current_setting('fsync') = 'off'
+  OR current_setting('wal_sync_method') IN ('open_sync', 'open_datasync')
   OR :io_sum_wal_normal_after_fsyncs > :io_sum_wal_normal_before_fsyncs;
  ?column? 
 ----------

src/test/regress/sql/stats.sql

@@ -672,6 +672,7 @@ SELECT sum(writes) AS writes, sum(fsyncs) AS fsyncs
 SELECT current_setting('synchronous_commit') = 'on';
 SELECT :io_sum_wal_normal_after_writes > :io_sum_wal_normal_before_writes;
 SELECT current_setting('fsync') = 'off'
+  OR current_setting('wal_sync_method') IN ('open_sync', 'open_datasync')
   OR :io_sum_wal_normal_after_fsyncs > :io_sum_wal_normal_before_fsyncs;
 
 -- Change the tablespace so that the table is rewritten directly, then SELECT