Add a std IO protocol interface

[nyonson]

Add a std library I/O (aka blocking) high level Protocol interface. Also experiments with some new API refactors which could be good for the existing async version.

[rustaceanrob]

Both of these will panic for an empty payload. Not sure if that is possible or not, but pointing it out for reference.

[nyonson]

Oops good point, shoulda thought more about this change.

[rustaceanrob]

Overall looks good and I like the new flow. Would just have to think for a while if someone could somehow send an empty payload contents, in which case contents would panic.

[nyonson]

Would just have to think for a while if someone could somehow send an empty payload contents, in which case contents would panic.

I think it would be difficult to not send a header byte, but still feels like this should be more defensive.

[rustaceanrob]

Could someone just send [0, 0, 0] as the length field? I need to actually re-read the spec but that was my gut reaction

[nyonson]

Could someone just send [0, 0, 0] as the length field? I need to actually re-read the spec but that was my gut reaction

I think that is against the spec because a header byte and a 16 byte tag are required. But a bug in a client or a bad actor could send it anyways. Right now, that case specifcally is handled by this impl in the low level decrypt function with a CiphertextTooSmall error, we check that at least a tag (16 bytes) is in the ciphertext. But maybe that should be bumped to 17 since there must also be a header byte to tell us if it is a decoy or not.

[rustaceanrob]

Interesting we didn't run into this scenario before. I think everything checks out though. All cases abuse should now lead to an EOF if I'm understanding correctly.