feat(extractor): implement multi-engine article content extractor

- Added scripts/extract_article_contents.py for batch scraping with stealth Foxcape and triple extraction (Trafilatura, Newspaper4k, Readability)
- Created unit, integration, and E2E test suite in tests/test_extract_article_contents.py (90/90 passing)
- Updated specs/003-article-content-extractor and README.md with usage documentation and CLI contracts
- Passed ruff linting/formatting and mypy type checking cleanly

tests/test_benchmark_24.py

@@ -6,19 +6,17 @@ Target Success Criterion: Precision >= 90% over the 24 cases.
 
 import json
 from pathlib import Path
+
 import pytest
-from src.models import ECPSnapshot, DecisionCategory
+
 from src.classifier import InherenceClassifier
+from src.models import ECPSnapshot
 
 FIXTURES_DIR = Path(__file__).parent / "fixtures" / "benchmark_24"
 LANGUAGES = ["pt", "en", "es", "de", "it", "fr"]
 DECISION_TYPES = ["direct", "contextual", "tangential", "not_related"]
 
-BENCHMARK_CASES = [
-    (lang, dec_type)
-    for lang in LANGUAGES
-    for dec_type in DECISION_TYPES
-]
+BENCHMARK_CASES = [(lang, dec_type) for lang in LANGUAGES for dec_type in DECISION_TYPES]
 
 
 @pytest.fixture(scope="module")
@@ -44,28 +42,28 @@ def test_benchmark_case(classifier, lang: str, dec_type: str):
     result = classifier.classify(ecp, content)
 
     # 1. Decision category validation
-    assert (
-        result.decision.value == expected["expected_decision"]
-    ), f"[{lang.upper()} - {dec_type}] Expected {expected['expected_decision']}, got {result.decision.value}. Rationale: {result.rationale}"
+    assert result.decision.value == expected["expected_decision"], (
+        f"[{lang.upper()} - {dec_type}] Expected {expected['expected_decision']}, got {result.decision.value}. Rationale: {result.rationale}"
+    )
 
     # 2. Derived is_inherent boolean validation
-    assert (
-        result.is_inherent == expected["expected_is_inherent"]
-    ), f"[{lang.upper()} - {dec_type}] Expected is_inherent={expected['expected_is_inherent']}, got {result.is_inherent}"
+    assert result.is_inherent == expected["expected_is_inherent"], (
+        f"[{lang.upper()} - {dec_type}] Expected is_inherent={expected['expected_is_inherent']}, got {result.is_inherent}"
+    )
 
     # 3. Language detection validation
-    assert (
-        result.detected_language == expected["expected_language"]
-    ), f"[{lang.upper()} - {dec_type}] Expected language '{expected['expected_language']}', got '{result.detected_language}'"
+    assert result.detected_language == expected["expected_language"], (
+        f"[{lang.upper()} - {dec_type}] Expected language '{expected['expected_language']}', got '{result.detected_language}'"
+    )
 
     # 4. Confidence threshold validation
     min_conf = expected.get("min_confidence", 0.0)
-    assert (
-        result.confidence >= min_conf
-    ), f"[{lang.upper()} - {dec_type}] Expected confidence >= {min_conf}, got {result.confidence}"
+    assert result.confidence >= min_conf, (
+        f"[{lang.upper()} - {dec_type}] Expected confidence >= {min_conf}, got {result.confidence}"
+    )
 
     # 5. Evidence presence for inherent content
     if result.is_inherent:
-        assert (
-            len(result.evidence) > 0
-        ), f"[{lang.upper()} - {dec_type}] Inherent decision must have non-empty evidence snippets"
+        assert len(result.evidence) > 0, (
+            f"[{lang.upper()} - {dec_type}] Inherent decision must have non-empty evidence snippets"
+        )