Ordering of columns not preserved in JDBC mapping vs table in external store [HZ-1556]

[sumnerib]

Describe the bug
When creating a JDBC SQL mapping without specifying the columns (like in below example), the column order is reversed compared to that in remote source table. This affects statements such as SELECT * FROM ... and INSERT INTO <mapping> VALUES.

CREATE MAPPING people
TYPE JDBC 
OPTIONS (
  'externalDataStoreRef'='mysql-database' 
)

Expected behavior
A clear and concise description of what you expected to happen.

To Reproduce

Steps to reproduce the behavior:

1. Start a member with an external-data-store configured
2. Create a JDBC mapping referencing the data store. This should be done without provide any columns explicitly (like above).
3. Execute SELECT * FROM <jdbc-mapping> or INSERT INTO <jdbc-mapping> VALUES to observe the behavior

Additional context

A potential workaround is to specify the columns in the desired order when creating the mapping. For example:

CREATE MAPPING people (
 id INT, 
 name VARCHAR 
) 
TYPE JDBC 
OPTIONS ( 
 'externalDataStoreRef'='mysql-database'
)

[github-actions]

Internal Jira issue: HZ-1556

[viliam-durina]

The column order isn't exactly reversed, but is lost because we put the columns into a HashMap here:

hazelcast/hazelcast-sql/src/main/java/com/hazelcast/jet/sql/impl/connector/jdbc/JdbcSqlConnector.java

Lines 157 to 166 in ee8ce2f

	Map<String, DbField> fields = new HashMap<>();
	for (int i = 1; i <= metaData.getColumnCount(); i++) {
	String columnName = metaData.getColumnName(i);
	fields.put(columnName,
	new DbField(metaData.getColumnTypeName(i),
	columnName,
	pkColumns.contains(columnName)
	));
	}
	return fields;