Support header protection for Autocrypt header

[link2xt]

We want to eventually support Header Protection for as many headers as possible. Code is already restructured to allow easy modification of the header confidentiality policy in #5715 so protecting Autocrypt should be as easy as removing these two lines:
https://github.com/deltachat/deltachat-core-rust/blob/429c14ae0bad035fae3fac6bfa3add63fa94d70a/src/mimefactory.rs#L746-L747

However, receiver currently handles Autocrypt header even before decrypting the message, so existing Delta Chat releases ignore Autocrypt headers received anywhere other than in the unprotected part.

As a first step for Autocrypt header protection we need receivers to be able to handle protected Autocrypt headers. Not handling received protected Autocrypt headers is a bug. Sending Autocrypt header is outside of the scope of this issue, we need releases supporting reception first.

[iequidoo]

What is interesting, looking at https://docs.autocrypt.org/level1.html#header-injection-in-outbound-mail, i don't see any requirements for where the Autocrypt header should be placed. E.g. for Autocrypt-Gossip the standard says:

An Autocrypt MUA MAY include Autocrypt-Gossip headers in messages. These headers MUST be placed in the root MIME part of the encrypted message payload.

Not handling received protected Autocrypt headers is a bug.

Probably, because the standard doesn't say that the header may not be protected. And how much Autocrypt-capable software has this bug? Should Delta Chat wait for them to fix it?

And returning to Autocrypt-Gossip, i don't see any code in mimefactory.rs preventing injecting gossips in unencrypted emails. Grepping by should_do_gossip, i see that gossips may be sent in groups where we don't encrypt because e.g. the majority of members prefer not to encrypt. So we reveal others' public keys to the network which is even worse than revealing our own key.

EDIT: I was wrong, the code adding gossip headers is under the let outer_message = if is_encrypted { condition above, so there's no bug.