test: Fix failing datetime tests in MariaDB

[bog-walk]

The following tests in exposed-java-time and exposed-kotlin-datetime fail when run using MariaDB:

KotlinTimeTests/'test string LocalDateTime with nanos'()
JavaTimeTests/'test string LocalDateTime with nanos'()

• Fails with the message Failed on microseconds because Exposed's datetime assertion functions for the fractional part are using roundToMicro() to compare nanoseconds. In most other DB, including MySQL, the nanoseconds retrieved from the DB is round to match the internal precision using RoundingMode.HALF_UP.
• This doesn't work because MariaDB always rounds down. For example, if a LocalDateTime with 815_716_723 (nanoseconds) is sent to the DB, the value retrieved from the DB is 815_716_000, not 815_717_000 as the tests expect.
• These tests are also flaky because they rely on Clock.System.now() (and the Java equivalent), which has an unknown nanosecond value, and then add a set amount of nanoseconds. This means any resulting sum value that happens to be on the lower half of a microsecond will be round down and pass the test. For example, 923_174_032 will correctly match with retrieved value 923_174_000.
• The assertion has been changed to floor the nanoseconds value instead and the test now inserts 2 values with constant nanoseconds to evaluate what happens when the value is on either side of the rounding half-way point.

Additional:

• Clean up assertion function: remove duplicate calls to currentDialectTest; remove unreachable branch (as PostgreSQLNG delegates to PostrgeSQL).

(Both modules)
MiscTableTest/testInsert01()
MiscTableTest/testInsert02()
MiscTableTest/testInsert03()
MiscTableTest/testInsert04()
MiscTableTest/testSelect01()
MiscTableTest/testSelect02()

• These also Failed on microseconds and are all fixed by the change above.