'str' object has no attribute 'policy' crash in BytesParser for certain .olk15Message files #5

Bug

A single .olk15Message file causes extraction to crash with 'str' object has no attribute 'policy' error.

Evidence

ERROR	messages	5B9C093F-3450-46F7-A067-FDEC803B6FCA.olk15Message	'str' object has no attribute 'policy'

Root Cause

writer.py:27 calls BytesParser(policy=email.policy.compat32).parsebytes(mime_bytes). The .olk15Message parser in parsers/message.py builds an EmailMessage and calls .as_bytes(), but the resulting MIME output contains something that Python's email parser chokes on — likely a malformed header or encoding issue where a header value is a str object instead of the expected type.

The _clean_text() function in parsers/message.py removes nulls but may not catch all binary artifacts that end up in header values.

Proposed Fix

1. In writer.py:27, wrap BytesParser.parsebytes() in a try/except with a fallback:
   • First try policy=email.policy.compat32
   • On failure, try writing the raw mime_bytes directly without re-parsing (since the bytes were already valid enough to be produced by EmailMessage.as_bytes())
2. In parsers/message.py, add a _sanitize_header() helper that ensures all msg[key] assignments produce clean, ASCII-safe strings before calling .as_bytes()
3. Log a warning when fallback is used so the user knows a message was written with reduced parsing