Adding optional code for automl train + prediction in one pipeline

[ScoGroMSFT]

• Adding optional code for demonstrating how to add prediction to an automl training pipeline
• Demonstrates the need to move automlpipelinebuilder-generated files between calls to avoid overwriting them.
• This was implemented on a customer engagement where they wanted to train + predict within the same pipeline
• The resulting pipeline looks like this: