Expensive subqueries may be evaluated during optimization in merge_key_fields

The following example extracted from https://mariadb.atlassian.net/browse/MDEV-430 results in evaluation of expensive subqueries during query optimization.

SET optimizer_switch = 'materialization=on,semijoin=on';

CREATE TABLE t1 (a INT, KEY(a)) ENGINE=MyISAM;
INSERT INTO t1 VALUES (8),(0);

CREATE TABLE t2 (b INT, c VARCHAR(1)) ENGINE=MyISAM;
INSERT INTO t2 VALUES (4,'j'),(6,'v');

CREATE TABLE t3 (d VARCHAR(1)) ENGINE=MyISAM;
INSERT INTO t3 VALUES ('b'),('c');

EXPLAIN
SELECT * FROM t1 WHERE a = (
SELECT MAX(b) FROM t2 WHERE c IN ( SELECT MAX(d) FROM t3 )
) OR a = 10;

The relevant call stack is:

#6 0x084866e9 in Item_singlerow_subselect::val_int (this=0x966d438) at /home/psergey/dev2/5.5-look7/sql/item_subselect.cc:1196
#7 0x0819053c in Item::update_null_value (this=0x966d438) at /home/psergey/dev2/5.5-look7/sql/item.h:1055
#8 0x08493a94 in Item_subselect::is_null (this=0x966d438) at /home/psergey/dev2/5.5-look7/sql/item_subselect.h:172
#9 0x082879fd in merge_key_fields (start=0x970e410, new_fields=0x970e42c, end=0x970e448, and_level=3) at /home/psergey/dev2/5.5-look7/sql/sql_select.cc:3874