[YSQL] Lack of failures on rapid row lock queries

[jaki]

Jira Link: DB-2190
When rapidly executing statements interleaved in transactions and related to row locking, there may be cases where failures don't happen when they should. For instance, executing the following rapidly can work (ignoring issue #2857):

Session A	Session B
CREATE TABLE t (i int, j int PRIMARY KEY)
INSERT INTO t VALUES (1, 10), (2, 20), (3, 30)
BEGIN ISOLATION LEVEL REPEATABLE READ
SELECT '(setting read point)'
	BEGIN ISOLATION LEVEL REPEATABLE READ
	SELECT '(setting read point)'
SELECT * FROM t FOR UPDATE
	SELECT * FROM t FOR UPDATE
COMMIT
	COMMIT

The test function PgMiniTest::TestRowLockConflictMatrix encounters this, so some time between queries is added as a workaround.

[pkj415]

I think the issue doesn't exist anymore since I don't see a sleep anymore between explicit row-locking statements in TestRowLockConflictMatrix().