Fix mimebody DKIM body-hash computation

[birktj]

This PR fixes the DKIM body-hash computation of singlepart and multipart bodies.

I opted for a fn body_raw(&self) -> Vec<u8> method for these bodies, but if it would be preferable with a fn format_body(&self, out: &mut Vec<u8>) instead I could do that change.

This fixes #922

[paolobarbolini]

Thanks for the PR, and sorry for the delay in reviewing. I don't mind the extra allocations, although it shouldn't be too much hard to remove them. Can you share the document that you used to determine that this is the right way of signing the bodies?

[birktj]

Thanks for having a look.

There shouldn't be have been any extra allocations the way the code was, but there could be in the future if body_raw ends up being used for something else. A format_body function would also reduce code duplication a bit as it could be reused in MultiPart::format and SinglePart::format. Thinking more about it I think it is a nicer solution so I have implemented this change now.

I am pretty sure that I used RFC 6376 as a reference, but note that this code does the exact same as the existing code, just making sure to not include the multipart headers when computing the body hash.

[paolobarbolini]

LGTM