scripts: build and deploy PDF pages for translations

[kbdharun]

• The page(s) follow the content guidelines.
• The PR title conforms to the recommended templates.

Changes

Add support for PDFs in other languages

• Add output flag to render.py, by default the filename without output flag is tldr-book.pdf (output flag allows specifying custom names, which I have used in CI).
• Update pt serif font to Noto Sans (which has support for almost all typefaces and 800+ languages, and is compatible with our license).
• Update CI to generate PDFs for individual translations (except a few languages, more information below).
• Update the deploy script to publish the translation PDFs.
• Update scripts/PDF README to document command changes.

Notes

I have tested all locales and verified rendering and edge cases.

The following locales ja, ko, zh, and zh_TW don't have a working render (even though the typeface is present in Noto) due to some issue in the weasyprint package. [Output of command:
missing_locales.txt for debugging]

Edit. Excluded some additional languages see #10846 (comment) for more information.

I have used the basic theme instead of the solarized-light (used on the English page) theme as it made reading certain typefaces easier.

Sample generated files in different themes

This was done during the PRs creation. Since then Tamil has been added to the exclusion list.

Tamil

tldr-pages-ta.pdf - Basic
tldr-pages-ta.pdf - Solarized light
tldr-pages-ta.pdf - Solarized dark

Related Issues

Closes #9432

[kbdharun]

Note: I am testing a wildcard method for CI. If you guys notice, some changes in CI it is for testing in a live environment. Once done I will rebase the changes.

[kbdharun]

Great, now the actions succeed and the files are uploaded (https://github.com/tldr-pages/tldr/actions/runs/6420703581). There seems to be a small issue where some files (bn, ml, ta, th) aren't rendered will trigger a rerun to debug it.

Artifact of the first run: pdf-artifacts.zip

Edit: the issue still seems to persist for the above 4 languages (the typefaces are indeed part of Noto Sans and while testing locale in a venv offline the files are rendered properly, they don't seem to render when built in the CI), will add these to the exclusion list too.

Edit 2. Done, cleaned up the additions that I used for testing in 0a73eb4 (It should be processed soon and will be reflected here). Now the PR should be good to go.

Artifact of the fixed run: pdf-artifacts (2).zip

Performance Impact on Workflow time

From testing, I can conclude that we might notice 3-5 minutes of added time to our existing PR runs since PDFs are built individually for language files.

[kbdharun]

Update: Migrated the building PDF part to a script that you can run inside the pdf directory. Started a test CI run, let's see how it goes.

Edit. It's done https://github.com/tldr-pages/tldr/actions/runs/6427813298?pr=10846 🎉 , I noticed I forgot to rename it to tldr-book in languages, will do it in the next commit.

Edit 2. Done, the artifact of the build can be found here.

[kbdharun]

Merged the PR, there seems to be an issue with the deploy step.

+ rm -rf /home/runner/work/tldr/tldr/language_archives
+ cp -f /home/runner/work/tldr/tldr/index.json /home/runner/site/assets/
+ cp -f '/home/runner/work/tldr/tldr/scripts/pdf/tldr-book*.pdf' /home/runner/site/assets/
cp: cannot stat '/home/runner/work/tldr/tldr/scripts/pdf/tldr-book*.pdf': No such file or directory
Error: Process completed with exit code 1.

Reverted the change, will debug this later today.