Add support for slowly changing mapping tables

## Background
Streaming conformance currently uses static mapping tables read from HDFS.
If a job is running long enough the mapping tables become outdated.

## Feature
Long-running streaming conformance jobs need a way to make in-memory mapping tables up to date.

Not sure yet how to pick up the changes to mapping tables.
- [ ] Periodically check the HDFS folder with mapping tables for updates
- [ ] Subscribe to a Kafka topic with changes to mapping tables

Which way to go depends on how MDM and MDR are done.

No file, test, or entry point is named. Start by locating the streaming conformance implementation and reading how MDM and MDR handle mapping-table changes, then decide whether HDFS polling or Kafka updates fits; done means long-running jobs keep their in-memory mapping tables current.

由索引模型根据 Issue 内容生成。