feat(enrichers): web_search DuckDuckGo + tests pytest de los 5 enrichers

Anade enricher web_search aplicable a nodos text/Concept/Topic. Hace
POST a html.duckduckgo.com con la query del nodo, parsea resultados
con HTMLParser stdlib, decodifica el redirect uddg= y crea N nodos
Url con relacion SEARCH_RESULT_OF apuntando al nodo origen.

Encadenable: tras web_search, fetch_webpage sobre cada Url completa
el pipeline search -> fetch -> extract.

Defensa contra ops_db_path mal resuelto: normaliza backslashes,
resuelve relativo contra app_dir, valida que la tabla entities
exista antes de tocar nada (exit codes 7/8/9 con JSON resumen).

Tests pytest (16/16 verde): conftest con operations.db temp +
schema minimo, stub de requests via PYTHONPATH para mockear red.
Cubre los 5 enrichers (extract_domain, fetch_webpage, extract_links,
extract_text_entities, web_search) + sanity check de manifests.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

tests/test_extract_links.py

@@ -0,0 +1,63 @@
+"""Tests del enricher extract_links — sin red, lee markdown del cache."""
+from __future__ import annotations
+
+from pathlib import Path
+
+from conftest import (
+    base_ctx, list_entities, list_relations, make_node, run_enricher,
+)
+
+
+SAMPLE_MD = """# Pagina demo
+
+Aqui hay [un enlace](https://example.com/articulo) interesante y
+otro [duplicado](https://example.com/articulo) que no debe contar
+dos veces.
+
+Tambien una URL pelada: https://otra.example/path?q=1
+y https://tercera.example/
+
+Y un email que NO debe extraer como Url: contact@no.example
+"""
+
+
+def test_extract_links_creates_url_nodes(ops_db, app_dir, registry_root):
+    # 1) Crear el cache con el markdown.
+    md_dir = Path(app_dir) / "cache" / "ab"
+    md_dir.mkdir(parents=True, exist_ok=True)
+    md_path = md_dir / "abc.md"
+    md_path.write_text(SAMPLE_MD, encoding="utf-8")
+    rel = md_path.relative_to(app_dir)
+
+    # 2) Crear Webpage con metadata.markdown_path apuntando al cache.
+    make_node(ops_db, node_id="w1", name="demo",
+              type_ref="Webpage", metadata={"markdown_path": str(rel)})
+
+    ctx = base_ctx(ops_db=ops_db, app_dir=app_dir, registry_root=registry_root,
+                   node_id="w1", node_name="demo", node_type="Webpage",
+                   metadata={"markdown_path": str(rel)})
+
+    rc, out, err = run_enricher("extract_links", ctx)
+    assert rc == 0, err
+    assert out is not None, err
+    assert out["entities_added"] >= 3, out
+
+    urls = [e["name"] for e in list_entities(ops_db, type_ref="Url")]
+    assert "https://example.com/articulo" in urls
+    assert "https://otra.example/path?q=1" in urls
+
+    rels = list_relations(ops_db, name="LINKS_TO")
+    assert len(rels) >= 3
+    assert all(r["from_entity"] == "w1" for r in rels)
+
+
+def test_extract_links_without_markdown_path_errors(ops_db, app_dir,
+                                                     registry_root):
+    make_node(ops_db, node_id="w1", name="demo",
+              type_ref="Webpage", metadata={})
+    ctx = base_ctx(ops_db=ops_db, app_dir=app_dir, registry_root=registry_root,
+                   node_id="w1", node_name="demo", node_type="Webpage")
+    rc, out, err = run_enricher("extract_links", ctx)
+    assert rc != 0, "deberia fallar sin markdown_path"
+    assert out is not None
+    assert "missing markdown_path" in (out.get("error") or "")