Allow override of the blob processing and create a subclass of JBossMQ's JDBC2 persistence manager
such that it can workaround the 4k limit of the thin driver.

Effectively we translate:
INSERT INTO JMS_MESSAGES (MESSAGEID, DESTINATION, TXID, TXOP, MESSAGEBLOB) VALUES (?,?,?,?,?)

into
INSERT INTO JMS_MESSAGES (MESSAGEID, DESTINATION, TXID, TXOP, MESSAGEBLOB) VALUES (?,?,?,?, EMPTY_BLOB())
SELECT MESSAGEID, MESSAGEBLOB FROM JMS_MESSAGES WHERE MESSAGEID = ? FOR UPDATE
UPDATE JMS_MESSAGES SET MESSAGEBLOB = ? WHERE MESSAGEID = ?

If we knew how to generically retrieve the primary key column and key values from the insert
we could do this as a plugin to the JDBC resource adapter.
But parsing SQL is not really an option.