The reason is that selecting from events_waits_history_long creates
a race condition: it can happen that SHOW EXPLAIN thread was kicked
off CPU exactly after posting a SHOW EXPLAIN request and then it wont
need to wait.
It doesn't seem to make sense to add more waits to stabilize the testcase.
Let's instead make a check that SHOW EXPLAIN statement has a
"stage/sql/show explain" stage.