Cache payload should be compressed before storing

Experiments show that an on disk cache compresses very well.

A 105MB cached payload compressed to 272KB with zip and 79KB with bzip2.

zip compression was much faster so seems like a good thing to implement.

The total cache size was 1.5GB and it came down to about 4MB after zip compression which is very manageable.

首先定位 Python 包的磁盘缓存写入和读取路径，然后检查缓存 payload 的序列化方式。为存储的 payload 实现 zip 压缩，并验证压缩条目仍可正确检索且缓存大小有所减少。

由索引模型根据 Issue 内容生成。