feat(quality): add L0 quality layer (Track A1 + A2 of dev plan)

L0 quality detection layer to catch translation failures BEFORE they
reach users. Pure Python/TypeScript, zero new dependencies, no API calls.

Backend (Python — services/quality/):
  - Script detection: 145 langs mapped to 23 scripts (Latin, Cyrillic,
    Greek, Arabic, Hebrew, CJK, Hangul, Kana, Devanagari, Bengali, etc.)
  - Language confusion detection (e.g. Arabic text for French target)
  - Arabic-script variant discrimination (Persian/Urdu/Pashto/Kurdish
    confusion — e.g. Persian text returned when Arabic was requested)
  - Length sanity check (with numeric/short-source exemptions)
  - Prompt leak detection (Translation: / Voici la traduction: / 翻译:)
  - Repetition hallucination detection (token + character level)
  - File text extraction for .docx/.xlsx/.pptx/.pdf (no translator
    changes needed)
  - Defensive pipeline that never raises (L0 must NEVER break a job)

Frontend (TypeScript — wordly.art---traduction-de-documents/src/utils/):
  - Exact 1:1 mirror of the Python module
  - Zero dependencies, works in browser AND Node.js
  - Native Unicode regex (\\p{L}/u) and codePoint iteration
  - 63 tests using Node's built-in test runner

Integration:
  - Feature-flagged: QUALITY_L0_ENABLED=false (default)
  - Observation only: logs structured events, never modifies files
  - try/except wrapped: impossible to break a translation job
  - Lazy imports: only loaded when flag is on
  - Zero impact on existing tests / behavior

Tests:
  - 111 Python tests covering all paths (config, script, length, leak,
    pipeline, file_extractor) — 100% pass
  - 63 TypeScript tests (Node --test) — 100% pass
  - 174/174 total tests for the L0 layer

Bug fixes in script mapping:
  - yi (Yiddish) -> hebrew (was incorrectly mapped to arabic)
  - dv (Maldivian) -> thaana (was incorrectly mapped to arabic)
  - ja (Japanese) -> hiragana_katakana (distinguishes from Chinese CJK)

Phase 1 (backend) + Phase 2 (frontend) of Track A complete.
Next: Track B1 (Word/Excel format preservation quick wins).

Closes Track A phase 1+2 of the dev plan.

tests/services/quality/test_file_extractor.py

@@ -0,0 +1,181 @@
+"""
+Tests for services/quality/file_extractor.py
+Uses real files generated via temporary paths.
+
+On Windows, tempfile.NamedTemporaryFile holds the file open and blocks
+overwrite, so we use a plain temp directory + manual filename instead.
+"""
+import tempfile
+import os
+from pathlib import Path
+
+import pytest
+
+from services.quality.file_extractor import (
+    extract_sample,
+    DEFAULT_MAX_SAMPLES,
+)
+
+
+class TestExtractSample:
+    def test_returns_empty_for_missing_file(self):
+        result = extract_sample(Path("/nonexistent/path/file.docx"), ".docx")
+        assert result == []
+
+    def test_returns_empty_for_none_path(self):
+        result = extract_sample(None, ".docx")
+        assert result == []
+
+    def test_returns_empty_for_unsupported_extension(self):
+        # .txt isn't supported — should return [] silently
+        with tempfile.TemporaryDirectory() as d:
+            p = Path(d) / "test.txt"
+            p.write_text("some text")
+            result = extract_sample(p, ".txt")
+            assert result == []
+
+
+def _make_tmp_path(suffix: str) -> Path:
+    """Create a unique temp file path. File does not exist yet."""
+    fd, name = tempfile.mkstemp(suffix=suffix)
+    os.close(fd)
+    return Path(name)
+
+
+class TestDocxExtraction:
+    def test_extracts_paragraphs(self):
+        from docx import Document
+        doc = Document()
+        doc.add_paragraph("Bonjour le monde")
+        doc.add_paragraph("Comment allez-vous?")
+        doc.add_paragraph("Merci beaucoup")
+        p = _make_tmp_path(".docx")
+        try:
+            doc.save(str(p))
+            result = extract_sample(p, ".docx", max_samples=10)
+            assert len(result) >= 1
+            texts = [s["translated"] for s in result]
+            assert "Bonjour le monde" in texts
+        finally:
+            try:
+                p.unlink()
+            except (OSError, PermissionError):
+                pass
+
+    def test_respects_max_samples(self):
+        from docx import Document
+        doc = Document()
+        for i in range(50):
+            doc.add_paragraph(f"Paragraphe numero {i} avec du texte")
+        p = _make_tmp_path(".docx")
+        try:
+            doc.save(str(p))
+            result = extract_sample(p, ".docx", max_samples=5)
+            assert len(result) == 5
+        finally:
+            try:
+                p.unlink()
+            except (OSError, PermissionError):
+                pass
+
+    def test_handles_empty_doc(self):
+        from docx import Document
+        doc = Document()
+        p = _make_tmp_path(".docx")
+        try:
+            doc.save(str(p))
+            result = extract_sample(p, ".docx")
+            assert result == []
+        finally:
+            try:
+                p.unlink()
+            except (OSError, PermissionError):
+                pass
+
+
+class TestXlsxExtraction:
+    def test_extracts_cells(self):
+        from openpyxl import Workbook
+        wb = Workbook()
+        ws = wb.active
+        ws["A1"] = "Bonjour"
+        ws["A2"] = "Monde"
+        ws["A3"] = "Comment"
+        p = _make_tmp_path(".xlsx")
+        try:
+            wb.save(str(p))
+            wb.close()
+            result = extract_sample(p, ".xlsx", max_samples=10)
+            assert len(result) >= 1
+            texts = [s["translated"] for s in result]
+            assert "Bonjour" in texts
+        finally:
+            try:
+                p.unlink()
+            except (OSError, PermissionError):
+                pass
+
+    def test_skips_numeric_cells(self):
+        from openpyxl import Workbook
+        wb = Workbook()
+        ws = wb.active
+        ws["A1"] = 100
+        ws["A2"] = 200
+        ws["A3"] = "Hello"
+        p = _make_tmp_path(".xlsx")
+        try:
+            wb.save(str(p))
+            wb.close()
+            result = extract_sample(p, ".xlsx")
+            texts = [s["translated"] for s in result]
+            assert "Hello" in texts
+            assert 100 not in texts  # numeric only cells skipped
+        finally:
+            try:
+                p.unlink()
+            except (OSError, PermissionError):
+                pass
+
+
+class TestPptxExtraction:
+    def test_extracts_slide_text(self):
+        from pptx import Presentation
+        pres = Presentation()
+        slide = pres.slides.add_slide(pres.slide_layouts[0])
+        slide.shapes.title.text = "Bonjour le monde"
+        p = _make_tmp_path(".pptx")
+        try:
+            pres.save(str(p))
+            result = extract_sample(p, ".pptx")
+            assert len(result) >= 1
+            assert result[0]["translated"] == "Bonjour le monde"
+        finally:
+            try:
+                p.unlink()
+            except (OSError, PermissionError):
+                pass
+
+
+class TestPdfExtraction:
+    def test_extracts_pdf_text(self):
+        # Create a minimal PDF using reportlab if available, else skip
+        pytest.importorskip("fitz")
+        try:
+            from reportlab.pdfgen import canvas
+        except ImportError:
+            pytest.skip("reportlab not available")
+        p = _make_tmp_path(".pdf")
+        try:
+            c = canvas.Canvas(str(p))
+            c.drawString(100, 750, "Bonjour le monde")
+            c.drawString(100, 700, "Comment allez vous")
+            c.save()
+            result = extract_sample(p, ".pdf")
+            assert len(result) >= 1
+            texts = [s["translated"] for s in result]
+            assert any("Bonjour" in t for t in texts)
+        finally:
+            try:
+                p.unlink()
+            except (OSError, PermissionError):
+                pass