Break decision-tree source into multiple components

[serendipious]

• Break plain objects into classes
• Move methods into separate modules
• Use classes to parse and validate training and test data-tables