Backport fix: prevent PDF image extraction from being skipped

[TheAryan77]

🧪 Add regression test for PDF image extraction (PyMuPDF)

Summary

Adds a regression test to ensure that extract_images=True does not silently skip
images when using PyMuPDFParser.

In langchain-community==0.4.1, image extraction was broken due to a premature
empty-buffer check, causing all images to be skipped with no warning. This test
would have failed in that version.

Why this matters

• The bug failed silently (no error or warning)
• OCR-based PDF pipelines produced incomplete results
• Difficult for users to diagnose without inspecting source code

This test ensures the expected behavior remains intact and prevents future
regressions.

Notes

• The underlying fix already exists on main
• This PR adds a regression test to lock in correct behavior