some refactors to make the code more readable

[JayceFayne]

The goal of this PR was to parse embedded rfc822 messages (eagerly).
For example:

From: sender@example.com
To: recipient@example.com
Subject: Wrapper message
MIME-Version: 1.0
Content-Type: message/rfc822

From: sender@example.com
To: recipient@example.com
Subject: Embedded message
MIME-Version: 1.0
Content-Type: text/plain

Hi

Is not being parsed but instead you get a binary body...which is fine.
It turns out adding this feature would just add unneccesarry complexity. I am opening this PR to upstream some refactors that make the parse_mail_recursive function much more readable.

[staktrace]

Thanks! Overall this is a nice cleanup. I'd prefer to drop the second commit but the other three look good

[staktrace]

I think I'd prefer to keep it as raw_data here. Either way it's inconsistent because the public function parse_mail has a parameter raw_data which is passed into this function, and the public struct ParsedMail has a field raw_bytes which is populated from raw_data. So no matter which one you use it's going to be inconsistent one way or the other. However if you think of this library as a data processor, with "raw data" coming and "raw bytes" going out, then this function is still before the conversion, so conceptually is more "raw data" than "raw bytes".