Versioning support for downstream libraries 💭

[Nolski]

Summary

The web crawler will overwrite the static json files it outputs on each crawl. We should have the json files named & stored using the versions of the associated documentation page.

Background

Describe the ideal solution:

Our crawler should explicitly define which versions of PyTorch, Tensorflow, and Tensorflow.js we want to support, and crawl those documentation pages explicitly. It should then note the version in the name of the file.

Outcome

This will allow us to be explicit about which versions of translation we support and not have any undocumented behavior.

[Nolski]

Resolved by #8

[Nolski]

Unfortunately we introduced a regression that reopened this issue.