Improve PANGAEA Publication Extraction Logic

[doswal]

fixes #51

Summary

This PR refactors the get_publications() workflow for PangaeaDataset to improve accuracy, robustness, and alignment with the PANGAEA metadata model. It replaces unreliable regex-based parsing with structured metadata access.

Key Changes

• Structured extraction

  • Use PanDataSet.supplement_to and PanDataSet.relations instead of regex parsing.

• Dataset citation handling

  • Use PanDataSet.citation directly (no Crossref).
  • Always included with Type = "citation".

• Selective Crossref usage

  • Applied only to:
    • supplement_to["uri"]
    • relations[i]["uri"]
  • Avoids failures for PANGAEA dataset DOIs.

• BibTeX correctness

  • Fixed conversion from bibtexparser entries to pybtex.Entry.
  • Proper handling of authors via Person objects.

• Deduplication

  • DOI-based deduplication within each study.

• Test stability

  • Added mocking for Crossref:

    @patch("doi2bib.crossref.get_bib")

  • Prevents conflicts with existing requests.get mocks.

Behavior

Source	Type
Dataset citation	"citation"
supplement_to	"supplement to"
relations	relation type
Regex parsing	removed

Motivation

• Regex-based extraction was fragile and inaccurate
• Crossref is unreliable for dataset DOIs
• PANGAEA provides structured relationships natively

This change improves correctness, maintainability, and testability.