Flink: Introduce IcebergCatalogLoader for loading table at runtime

[JingsongLi]

Unlike Spark catalog table (Table is only required on the client/driver side), Flink needs obtain Table object in Job Manager or Task.

• For writer (Flink: Add the iceberg files committer to collect data files and commit to iceberg table. #1185): Flink needs obtain Table in committer task for appending files.
• For reader (Flink: Implement Flink InputFormat and integrate it to FlinkCatalog #1293): Flink needs obtain Table in Job Manager for planing tasks.

So we can introduce a CatalogLoader for reader and writer, users can define a custom catalog loader in FlinkCatalogFactory.

public interface CatalogLoader extends Serializable {
  Catalog loadCatalog(Configuration hadoopConf);
}