Encoding relations API

[sgwilym]

Willow's site will soon define encoding relations for many encodings, which are a set of acceptable encodings for a given type. These will be useful for things like WGPS messages where we are decoding many messages and we do not want the implicit overhead of verifying that every message was encoded correctly.

We still want to retain the concept of canonical encodings, as these are important for things like Meadowcap signatures.

So this PR adds new RelationDecodable and RelativeRelationDecodable (open to suggestions) traits with a decode_relation and relative_decode_relation methods on them, respectively.

The methods have a default implementation which is an alias for Decodable / RelativeDecodable's decode_canonical and relative_decode_canonical.

I made these traits separate so that we can define types which have only a canonical decoder, or only a relation decoder (WGPS messages), or both!

Fuzz tests have been refactored to be synchronous and inclued additional checks on encoding tests.

There is also a new shell script for running all the fuzz tests sequentially (see the README) (closes #54).

[AljoschaMeyer]

Looks pretty good, just some small suggestions.

[AljoschaMeyer]

How do we feel about not encoding with encoding relations?

I feel very good about that.

[sgwilym]

As I'm working through these encodings, I think I'm going to split off relation decoding into its own trait.

My motivation for this would be to make it possible to implement types without canonical decoders which check for errors.

For example, it would be a huge waste of time to implement canonical decoders for all of the WGPS message types, when we will only ever decode relations.

Thoughts @AljoschaMeyer?

[AljoschaMeyer]

Sounds correct.