feat(ocr): increment ocr_skipped_pages_total on per-page engine failure

Bumps the counter in both /ocr/stream except blocks (standard and guided
generators) so the existing skipped_pages local variable now also flows
into Prometheus.

Refs #652 (AC3b)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

ocr-service/test_metrics.py

@@ -187,3 +187,46 @@ async def test_ocr_pages_total_incremented_once_per_page_in_stream(fresh_metrics
 
     value = fresh_metrics.ocr_pages_total.labels(engine="surya")._value.get()
     assert value == 3.0
+
+
+@pytest.mark.asyncio
+async def test_ocr_skipped_pages_total_incremented_when_engine_raises_for_a_page(fresh_metrics):
+    """When the engine raises on a page, ocr_skipped_pages_total bumps and the stream finishes."""
+    mock_images = [Image.new("RGB", (100, 100)) for _ in range(2)]
+    good_blocks = [{"pageNumber": 1, "x": 0.0, "y": 0.0, "width": 1.0, "height": 1.0,
+                    "polygon": None, "text": "ok", "words": []}]
+
+    call_count = {"n": 0}
+
+    def extract_side_effect(*args, **kwargs):
+        call_count["n"] += 1
+        if call_count["n"] == 1:
+            raise RuntimeError("synthetic engine failure")
+        return good_blocks
+
+    with patch("main.kraken_engine.load_models"), \
+         patch("main.load_spell_checker"), \
+         patch("main._download_and_convert_pdf", new_callable=AsyncMock, return_value=mock_images), \
+         patch("main.preprocess_page", side_effect=lambda img: img), \
+         patch("main.surya_engine.extract_page_blocks", side_effect=extract_side_effect):
+        async with AsyncClient(transport=ASGITransport(app=app), base_url="http://test") as client:
+            import main as main_module
+            main_module._models_ready = True
+            try:
+                async with client.stream("POST", "/ocr/stream", json={
+                    "pdfUrl": "http://minio/doc.pdf",
+                    "scriptType": "TYPEWRITER",
+                    "language": "de",
+                }) as response:
+                    assert response.status_code == 200
+                    saw_error = False
+                    async for line in response.aiter_lines():
+                        if line and '"type": "error"' in line:
+                            saw_error = True
+                    assert saw_error
+            finally:
+                main_module._models_ready = False
+
+    assert fresh_metrics.ocr_skipped_pages_total._value.get() == 1.0
+    # The second page still succeeds.
+    assert fresh_metrics.ocr_pages_total.labels(engine="surya")._value.get() == 1.0