At a certain point when expecting a line terminator parts() checks for
'\n' but not '\r\n'. This resulted in all multipart attachments coming
from b"\r\n" sources like IMAP having only one part when parsed.
Attachment needs to know the range of bytes where the body part of the
attachment is located. The Attachment.raw field contains the entire
attachment, headers and body. The new Attachment.body fields contains a
`StrBuilder` which contains the offset and length of the body part inside
`raw`.
'\n'
when an empty header is last, the rest of the body keeps getting parsed
as headers. when header starts with '\n' because the value is long, the
value gets parsed as a name and the header parser fails.
closes#100closes#101closes#122