fix: re-raise exception in ingest_document so Celery retry mechanism fires (#561)

[ionfwsrijan]

Description

ingest_document() wraps its entire pipeline in a try/except block. The except Exception handler logs the error, rolls back the DB session, marks the document as failed, and commits. However, the exception is completely consumed — never re-raised.

Meanwhile, the Celery task process_document is decorated with autoretry_for=(Exception,), max_retries=3, acks_late=True — all designed to handle transient failures. But because ingest_document never propagates an exception, none of them ever fire.

Changes

backend/app/services/document_ingestion.py

• Added raise after the except block so the original exception propagates to Celery's autoretry_for after DB state is cleaned up (rollback, mark failed, commit).

backend/app/tasks.py

• Changed the return value to query the document's actual DB status after ingest_document returns, instead of hardcoding "completed". This ensures the task result accurately reflects success or failure.

Impact

• Transient failures (API timeout, DB hiccup) now trigger automatic Celery retries
• Task result status matches DB status consistently
• Framework-level retry configuration (max_retries=3, default_retry_delay=30) is now functional

Closes #561

[ionfwsrijan]

@param20h Please review this