Fix bugs not to discard statistics when changing stats_fetch_consistency

Hi, hackers

There is below description in docs for stats_fetch_consistency.
"Changing this parameter in a transaction discards the statistics 
snapshot."

However, I wonder if changes stats_fetch_consistency in a transaction, 
statistics is not discarded in some cases.

Example:
--
* session 1
=# SET stats_fetch_consistency TO snapshot;
=# BEGIN;
=*# SELECT wal_records, wal_fpi, wal_bytes FROM pg_stat_wal;
  wal_records | wal_fpi | wal_bytes
-------------+---------+-----------
        23592 |     628 |   5939027
(1 row)

* session 2
=# CREATE TABLE test (i int); -- generate WAL records
=# SELECT wal_records, wal_fpi, wal_bytes FROM pg_stat_wal;
  wal_records | wal_fpi | wal_bytes
-------------+---------+-----------
        23631 |     644 |   6023411
(1 row)

* session 1
=*# -- snapshot is not discarded, it is right
=*# SELECT wal_records, wal_fpi, wal_bytes FROM pg_stat_wal;
  wal_records | wal_fpi | wal_bytes
-------------+---------+-----------
        23592 |     628 |   5939027
(1 row)

=*# SET stats_fetch_consistency TO cache;

=*# -- snapshot is not discarded, it is not right
=*# SELECT wal_records, wal_fpi, wal_bytes FROM pg_stat_wal;
  wal_records | wal_fpi | wal_bytes
-------------+---------+-----------
        23592 |     628 |   5939027
(1 row)
--

I can see similar cases in pg_stat_archiver, pg_stat_bgwriter, 
pg_stat_checkpointer, pg_stat_io, and pg_stat_slru.
Is it a bug? I fixed it, and do you think?

-- 
Regards,
Shinya Kato
NTT DATA GROUP CORPORATION