fix(ocr): address code review feedback on multi-language OCR

- Fix _get_pipeline_ocr_language: remove redundant `or None` in step_config.get()
- Add Session type hint to _get_pipeline_ocr_language db parameter via TYPE_CHECKING
- Update process_with_ocr to use modern str | None syntax instead of Optional[str]
- Fix test_get_pipeline_ocr_language_explicit_pipeline_takes_priority: properly add
  sys_step to db_session so the system pipeline step is persisted in the test DB

Co-authored-by: christianlouis <361235+christianlouis@users.noreply.github.com>

app/tasks/process_with_ocr.py

@@ -17,7 +17,6 @@ task with a multi-engine OCR pipeline that:
 
 import logging
 import os
-from typing import Optional
 
 from app.celery_app import celery
 from app.config import settings
@@ -36,9 +35,9 @@ logger = logging.getLogger(__name__)
 def process_with_ocr(
     self,
     filename: str,
-    file_id: Optional[int] = None,
-    original_text: Optional[str] = None,
-    language: Optional[str] = None,
+    file_id: int | None = None,
+    original_text: str | None = None,
+    language: str | None = None,
 ):
     """Run the configured OCR providers on *filename* and continue the pipeline.