I don't like methods that do something completely different with a null parameter

I was using the putIfAbsent API wrong

just seen a ConcurrentModificationException on Jenkins.

It looks like we can access the Map<Integer, Chunk> from different threads concurrently without taking the same locks. So make it a ConcurrentHashMap. Nothing useful we can lock on here without introducing an ABBA deadlock.

now that we require Java1.7

ResultSet#getObject is missing some conversions in table B.4 of the
JDBC specification.

- add support for Blob in getObject
- add support for Clob in getObject
- add support BigInteger <-> BIGINT
- add support java.util.Date <-> TIMESTAMP
- add support java.util.Calendar <-> TIMESTAMP
- add tests for SMALLINT
- add tests for BIGINT

where we are calling getException() and seeing a null value.
Because we access these without synchronization, we need to use volatile.

shows up as random failures in the TestMVStore.testFastDelete test case