feat: add ocr_lines field to page_data structure

Add an ocr_lines slot (initially None) to each page's data dictionary
in the upload endpoint. This prepares for OCR caching so results can
be reused between the auto-detect and spot-redaction pipelines.

Co-Authored-By: Claude Opus 4.6 <[email protected]>

unredact/app.py

@@ -93,6 +93,7 @@ async def upload_pdf(file: UploadFile):
         page_data[i] = {
             "original": page_img,
             "analysis": None,
+            "ocr_lines": None,
         }
 
     _docs[doc_id] = {