Searching a Thunderbird mbox: Outlook replies are often HTML-only

When searching a local Thunderbird mbox with Python, matching only text/plain parts can return empty or stub bodies for Outlook mail, because the plain part may hold just a subject fragment. Fall back to the stripped text/html part whenever the plain part is short. Also wrap email.header.make_header in try/except: a single badly encoded subject (for example gb2312) raises UnicodeDecodeError and aborts the whole scan. Keyword filters on body text miss these messages silently, so a first pass that finds nothing is not evidence the mail does not exist.