Accelerate CI by commonizing thrift build and leveraging cache

Building thrift compiler takes several minutes and occupies approximately one out of four or five in the total CI duration.
It can be improved by optimizing workflow and leveraging cache so as to avoid building thrift many times.

**Reporter**: [Kengo Seki](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=sekikn) / @sekikn
**Assignee**: [Kengo Seki](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=sekikn) / @sekikn

_{**Note**: *This issue was originally created as [PARQUET-2205](https://issues.apache.org/jira/browse/PARQUET-2205). Please see the [migration documentation](https://issues.apache.org/jira/browse/PARQUET-2502) for further details.*}

首先检查 repository 的 CI 工作流配置，以及 Thrift 编译器在各个 job 中的构建方式；找出重复构建和现有的缓存边界。当编译器被统一使用或缓存，使 CI 避免不必要的重复构建并缩短 CI 的总体时长时，即视为完成。

由索引模型根据 Issue 内容生成。