fix(parser): bare except Exception swallows MemoryError during per-page parse -- process continues in unstable state #295

Summary

The per-page try/except in parser.py at line ~161 uses bare 'except Exception as exc:' which catches MemoryError (a subclass of Exception in Python). A corrupt page that causes a MemoryError is silently swallowed and logged as a page-skip warning, potentially allowing the pipeline to continue in an unstable memory state.

Fix

ingestion/parser.py line ~161: narrow the except to expected fitz error types:
except (fitz.FitzError, RuntimeError, AttributeError, ValueError) as exc:
And either let MemoryError propagate naturally, or re-raise after logging.

Impact

A large corrupt PDF page that causes OOM is silently skipped. The process may continue in a low-memory state causing unpredictable failures in subsequent processing.