Add spec for binary token serialization

[Egge21M]

This adds some standardization for upcoming raw / binary token serialization, simply prepending the token with a identifier, as well as a version "byte" and encoding everything using cbor.

[prusnak]

I am not a big fan of this proposal because it requires cbor dependency for future versions which might not be based on cbor.

IMO much better encoding is just simple:

CRAW <uint8_t version> <binary data>

(where binary data is the data just before base64 encoding)

That way an app can very simply check whether it understands the payload by checking the first 5 bytes.

Also I think that we should use binary version 03 for v3 of token and 04 for v4 token.

[Egge21M]

I am not a big fan of this proposal because it requires cbor dependency for future versions which might not be based on cbor.

IMO much better encoding is just simple:

CRAW <uint8_t version> <binary data>

That way an app can very simply check whether it understands the payload by checking the first 5 bytes.

Very good point, however this is meant to be for transports that are binary only. So we should probably do something like

utf8(CRAW) <uint8_t version> <binary data>

[prusnak]

ACK

[callebtc]

LGTM