[Java] Improve testing of Dataset bindings

From https://github.com/apache/arrow/pull/13811

- We should ensure all types are tested
- We should organize tests in a way that Parquet, IPC, and eventually CSV/ORC can mostly share test code (save for perhaps skipping/overriding specific format-type pairs)

Incidentally: it may be good to incrementally port this module to JUnit5 and drop JUnit4

**Reporter**: [David Li](https://issues.apache.org/jira/browse/ARROW-17342) / @lidavidm

_{**Note**: *This issue was originally created as [ARROW-17342](https://issues.apache.org/jira/browse/ARROW-17342). Please see the [migration documentation](https://github.com/apache/arrow/issues/14542) for further details.*}

首先定位 Dataset bindings 的 Java 測試，並檢視 Parquet 和 IPC 案例目前是如何組織的。檢查所有型別的涵蓋情況，然後評估測試如何在不同格式之間共用程式碼，包括可能的 CSV 和 ORC 案例；JUnit4 到 JUnit5 的遷移被列為次要考量。

由索引模型根據 Issue 內容生成。