Insert too-large double produces wrong exception via Postgres

Using the Postgres server to insert a too-large value produces a parse
exception rather than an overflow exception (error code 42000
vs. 55004).

Note that it is also unclear why this overflow should produce an
exception when integer overflows are truncated.

SQL:

  CREATE TABLE t (double_field double)
  INSERT INTO t VALUES (1.7976931348623157E309)

Result:

  org.postgresql.util.PSQLException: ERROR: [] Floating point exponent overflow: INSERT INTO t VALUES (1.7976931348623157E309)
 at org.postgresql.core.v3.QueryExecutorImpl.receiveErrorResponse(QueryExecutorImpl.java:2102)
 at org.postgresql.core.v3.QueryExecutorImpl.processResults(QueryExecutorImpl.java:1835)
 at org.postgresql.core.v3.QueryExecutorImpl.execute(QueryExecutorImpl.java:257)
 at org.postgresql.jdbc2.AbstractJdbc2Statement.execute(AbstractJdbc2Statement.java:500)
 at org.postgresql.jdbc2.AbstractJdbc2Statement.executeWithFlags(AbstractJdbc2Statement.java:374)
 at org.postgresql.jdbc2.AbstractJdbc2Statement.execute(AbstractJdbc2Statement.java:366)
 at com.akiban.sql.pg.YamlTester$StatementCommand.execute(YamlTester.java:436)
 ... 44 more

To reproduce:

  mvn test -Dtest=PostgresServerMiscYamlIT -Dcom.akiban.sql.pg.PostgresServerMiscYamlIT.CASE_NAME_REGEXP=bug-insert-double-overflow

Reproduced in trunk revision 1239