Feature request: Support for HTML / MHTML files (e.g. Jira exports)

## Problem

Jira exports issues as standalone HTML files wrapped with `Content-Type: application/vnd.ms-word`, often saved with a `.doc` extension. These files are plain HTML documents (not OLE2 compound binaries), and anydoc currently rejects them as malformed:

```
malformed document: not an OLE2 compound file: Invalid CFB file (wrong magic number): [3c, 21, 44, 4f, 43, 54, 59, 50]
```

The magic bytes `3c 21 44 4f 43 54 59 50` = `` header (or `` root element) and routing through the existing HTML-to-model pipeline would cover the Jira export case. Full MHTML (multipart MIME with embedded resources like images) support would be a bonus.

## Example file

[jira_export_ticket.doc](https://github.com/user-attachments/files/30816559/AI1SOLHPCHAT-98.doc)

Trace the EPUB parser and the existing HTML-to-model pipeline first, then reproduce the failure with the linked Jira export. Define support around standalone HTML recognition and conversion, and verify the example file produces Markdown; treat full MHTML resource handling as a separate scope unless clarified.

Written by the indexing model from the issue text.