Force ScriptImplementor to explicitly name columns in SELECT

[jogrogan]

There is a regression in Flink 1.20, previously you could have two schemas where one is a subset of another. Take the following example:

Source Table:

CREATE TABLE IF NOT EXISTS `source` (`field1` VARCHAR, `field2` VARCHAR) WITH (...);

Sink Table:

CREATE TABLE IF NOT EXISTS `sink` (`field1` VARCHAR, `field2` VARCHAR, `field3` VARCHAR) WITH (...);

A flink SQL statement like the following would be valid (we would generate this prior to this PR):

INSERT INTO `sink` (`field1`, `field2`) SELECT * FROM `source`;

In Flink 1.20 this throws an exception with: "Number of columns must match number of query columns".

The fix is to explicitly reference fields.

INSERT INTO `sink` (`field1`, `field2`) SELECT `field1`, `field2` FROM `source`;

This PR changes the behavior to always explicitly name fields, even when a SELECT * would be valid.

[ryannedolan]

weird!