Multimodal PDF support

[jamesbraza]

This PR completes MVP multimodal support for PaperQA:

1. Moves PDF readers to support images and tables
   • With several routes such as full page screenshot vs image screenshot
2. Creates an opt-out setting for multimodal parsings
3. Adds test PDF checking we can parse table data

[Copilot]

Pull Request Overview

This PR implements MVP multimodal support for PaperQA, allowing the system to parse and utilize images and tables from PDFs alongside text content. The implementation introduces an opt-out setting for multimodal parsing and adds comprehensive support across both PyPDF and PyMuPDF parsers.

• Adds multimodal parsing capability to PDF readers supporting images, tables, and full-page screenshots
• Introduces multimodal configuration setting (defaults to True) to control image/table parsing
• Creates comprehensive test coverage for table querying and multimodal functionality

Reviewed Changes

Copilot reviewed 11 out of 16 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
src/paperqa/settings.py	Adds multimodal boolean field to ParsingSettings for controlling image/table parsing
src/paperqa/docs.py	Integrates multimodal setting into document reading pipeline via parse_images parameter
packages/paper-qa-pypdf/src/paperqa_pypdf/reader.py	Implements multimodal support with pypdfium2 for full-page screenshots and media parsing
packages/paper-qa-pymupdf/src/paperqa_pymupdf/reader.py	Extends PyMuPDF parser to support drawings, tables, and full-page screenshots with clustering
tests/test_paperqa.py	Updates existing tests for multimodal behavior and adds new table querying test
tests/test_agents.py	Updates file count expectations to include new influence.pdf test file
Various test files	Comprehensive test coverage for multimodal parsing functionality across both PDF parsers

[mskarlin]

could we add some error handling here if a bad image is hit and returning an ImpossibleParsingError?

[jamesbraza]

We already protect page = file.load_page(...) with ImpossibleParsingError, I think once a Page is loaded in and constructed, we should be good.

I haven't seen a get_pixmap crash so far, I'd like to hold off on this for the scope of PR

[mskarlin]

lgtm -- some minor comments