Breaking changes in JSON encoders and decoders

[stil4m]

Breaking changes in JSON encoders and decoders

[stil4m]

I've just merged the PR ;). Had some changes I wanted to bring to master as well.

[stil4m]

@MartinSStewart I see the advantages on the reduction of the JSON data. However, I'm not sure if it has to be implemented directly. It is possible to rewrite the current implementation to the Codecs without breaking the format. And indeed it reduces the risk a bit, mainly on JSON field names that have to be written twice in normal Json.Decode|Encode usage.

Performance might be an issue. Unsure about this. I might do a POC on this.

[MartinSStewart]

Alright, I'll go ahead and fix the encoder/decoder for the pull request without elm-codec for now.

[miniBill]

As I've written in miniBill/elm-codec#4, I don't expect performance to be any different, except for the recursive codecs, where I have to build the pair "at runtime" instead of just building the single encoder/decoder I need.
For custom types I now use a Dict internally that, while not as fast as a case ... of shouldn't be bad