Description
Currently, all the on-chain datastructures use tuple representation (structs are serialized as an array of their fields in order), this is for space efficiency and performance.
However, we initially thought that it was fine to not use tuple representation for the cbor-rpc messages, as those dont go on chain, and its not that much extra space.
After some discussion with @dignifiedquire, it seems like the thing that makes the most sense, and leaves the fewest surprises is to also use tuple representation for the cbor-rpc data structures. This makes sense when you think about data structures that contain the same datastructures that you have on chain (blocks, messages, etc). It feels more natural to write a single way of serializing everything and just use it.