More work on 5.2 migrator.

Previous migrator  could fail with out of memory error, either in java itself or the database. The migrator 'batches' work to migrate processing all txns for a given contact in a single batch ( a batch might involve multiple contacts but each contact is processed in a single batch).  Problem is some systems, especially systems where we migrated data from competitor systems which defined cash sales as a contact, the # of txns for a single customer can be insanely large (Typically batch size is around 500 txns, but some systems can have customers with 70k txns against a single customer). New migrator includes some smarts to create much smaller batches of work, avoiding running into memory issues ( batch based on network graphs of interlinked txns, instead of batching by customer).