feat(storage): implement immutable original/processed file storage and collision handling

- Add original_file_path and processed_file_path columns to FileRecord model
- Create database migration for new fields
- Implement get_unique_filepath_with_counter() with -0001 suffix format
- Update process_document to save immutable copy to /workdir/original
- Add force_cloud_ocr parameter to process_document for forced OCR
- Update embed_metadata to use new collision handling
- Update metadata JSON to include file path references
- Add /files/{file_id}/reprocess-with-cloud-ocr API endpoint
- Update processed_file_path in database during embedding

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

migrations/versions/002_add_file_paths.py

@@ -0,0 +1,29 @@
+"""Add original_file_path and processed_file_path to FileRecord
+
+Revision ID: 002_add_file_paths
+Revises: 001_file_processing_steps
+Create Date: 2026-02-11
+
+"""
+
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "002_add_file_paths"
+down_revision: Union[str, None] = "001_file_processing_steps"
+depends_on: Union[str, None] = None
+
+
+def upgrade() -> None:
+    """Add original_file_path and processed_file_path columns to files table."""
+    op.add_column("files", sa.Column("original_file_path", sa.String(), nullable=True))
+    op.add_column("files", sa.Column("processed_file_path", sa.String(), nullable=True))
+
+
+def downgrade() -> None:
+    """Remove original_file_path and processed_file_path columns from files table."""
+    op.drop_column("files", "processed_file_path")
+    op.drop_column("files", "original_file_path")