Drop python-ecdsa and port all crypto operations to libsodium

[lsd-cat]

See here #23.

We might want to test some more before merging, especially functionalities outside the demo.sh, such as attachments.

[eaon]

This is great! 🎉 There's a couple of comments that it seems like you overlooked, but otherwise this is much easier to understand than before.

Other than that (I didn't make inline notes about it because it also isn't directly an issue of this PR) there's a couple of abbreviations that aren't easy to understand/infer the meaning of. gdh for example may be misinterpreted as Gap-DH, which we don't use. So just noting that before the imminent freeze, it might still make sense to try and find clearer terminology

[eaon]

Is this intended as a form of documentation or did you overlook this before committing it?

[lsd-cat]

Maybe it was the late night, but I lost a good half an hour debugging that out so I left it as a reminder, even though I agree it is poorly written

[lsd-cat]

Thanks @eaon I agree with uncommenting. I will push a couple more changes tomorrow dropping the journalist_uid code-wise, since Ed25519 public keys are already 32 bytes long, there is no purpose of using an hash as a key id rather than the key itself, plus it adds sha3 as a dependency in case we want to interact with other programming languages (I am doing some cross-language experiments with libsodium bindings).

[lsd-cat]

Following #23 (comment) we might want to rethink the usage of libsodium Box for source to journalist message encryption .

[eaon]

What about using SealedBox?

[lsd-cat]

Opting to merge this while we complete the protocol discussion here #30

[cfm]

I've reviewed this at a high level, to understand how the implementation changes from mixed ecdsa+nacl to all nacl. I'm happy to approve and merge it for ongoing evaluation of the proof of concept.