Enh: add support for mboxrd format

Body lines in mbox files must not start with the characters 'From ' as the string identifies the start of the next message.

Unfortunately there is no single standard for how such lines are handled.

The mboxo format changes 'From ' to '>From '; this is ambiguous when attempting to regenerate the original email.

The mboxrd format additionally changes '>From ' to '>>From ' etc, which is reversible.

Apache HTTPD mod_mbox currently uses mboxo (as far as I am aware).
The ASF mbox-vm archiver uses mboxrd.

PonyMail currently uses mboxo for generating mbox downloadfiles and defaults to reading mbox files as mboxo.

Whilst this is self-consistent, it is not ideal.
The code should allow for mboxrd support.

Start by tracing the code that reads mbox files and generates mbox download files, since the issue identifies both paths as using mboxo. Compare their current escaping and unescaping behavior with the mboxrd rules described here; done means mboxrd input and output are supported while existing mboxo behavior remains available.

Written by the indexing model from the issue text.