Add cache bundle bootstrap and improve abstract ingestion

[StevenZ904]

Summary

• add cached multi-source abstract ingestion and stronger DB rebuild behavior
• add cache bundle export/install support for bootstraping from a published snapshot
• update README with bundle bootstrap instructions and the published Drive link

Validation

• pytest -q
• clean-environment install, bundle download, bundle install, query, and incremental build smoke test

[Kyle-Kyle]

Hmm. So, the main purpose of the bundle is to avoid duplicate requests when updating the database? Seems reasonable to me. But I don't like the implementation.

But I have a few comments and questions:

• I don't think putting a bundle on Google drive is a good idea: people downloading from the same link may have different content if it is edited. This is especially problematic since we do unzip on this untrusted data and do file extraction logic by ourselves. A better approach will be hosting the bundle on GitHub release page.
• Is the code generated by AI? Some logic seems unnecessary and fragile. Personally, I don't like huge chunks of AI-generated code (AI-generated patches that can be verified by humans are fine). I have had a terrible experience with it (it can bury landmines in the code that are difficult to debug).
• And I'll suggest that you to implement one logic in each commit.

Anyway. I'll carefully review and rewrite the code so that it is human-readable when I have some more cycles.