fix: replace fixed 3s sleep with load event + 0.5s in get_md

[g-eoj]

wait_until="commit" + asyncio.sleep(3) was fragile — slow pages could still be loading, fast pages wasted time.

Switching to wait_until="load" (waits for HTML/CSS/images) + a 0.5s sleep for JS rendering. Most research sources are server-rendered so 0.5s is sufficient.

Also adds *.ipynb to .gitignore.

Closes #4

Greptile Summary

Improved page loading reliability in get_md() by switching from wait_until="commit" to wait_until="load" and reducing sleep time from 3s to 0.5s.

• Replaced fragile timing approach that could fail on slow pages or waste time on fast pages
• wait_until="load" waits for HTML/CSS/images before proceeding (more robust than "commit")
• 0.5s sleep is sufficient for JS rendering on server-rendered research sources
• Added *.ipynb to .gitignore for local notebook experiments

Confidence Score: 5/5

• This PR is safe to merge with no issues found
• The changes are well-reasoned and improve both reliability and performance. The switch to wait_until="load" is more robust than the previous "commit" approach, and the reduced sleep time (0.5s vs 3s) is appropriate for the stated use case of server-rendered pages. The .gitignore addition is standard practice. No logic, syntax, or security issues detected.
• No files require special attention

Important Files Changed

Filename	Overview
.gitignore	Added *.ipynb to ignore Jupyter notebooks - standard practice
api/librarian/_tools.py	Replaced fragile 3s fixed sleep with wait_until="load" + 0.5s for JS rendering - more reliable and efficient

_{Last reviewed commit: 8d728c1}