Rename internal types and channels from Batch to Partition for consistency with the data model.
Add a YAML config schema for migration jobs (source/target schema, table, primary key, options) and parse it at startup.
