harden domain architecture and add lazy Crossref+Unpaywall PDF enrichment with publications workspace refactor

tests/unit/test_doi_normalize.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+from app.services.domains.doi.normalize import first_doi_from_texts, normalize_doi
+
+
+def test_normalize_doi_extracts_and_lowercases() -> None:
+    value = "https://doi.org/10.48550/ARXIV.1412.6980"
+    assert normalize_doi(value) == "10.48550/arxiv.1412.6980"
+
+
+def test_first_doi_from_texts_prefers_first_match() -> None:
+    result = first_doi_from_texts(
+        "no doi here",
+        "venue text 10.1000/ABC-123",
+        "title with 10.9999/ignored",
+    )
+    assert result == "10.1000/abc-123"