Add Message.from_json

[aslakhellesoy]

This PR adds a new Message.from_json class method that turns a JSON string into a message object.

It addresses three shortcomings:

• Lower camel case field names (fooBarZap) in the JSON are converted to underscore (fooBarZap)
• Values for bytes fields are Base64 decoded unless they are already in "binary" format. This is because Protobuf's JSON representation uses Base64 encoding for bytes fields.
• Enum values in the JSON can be either string or number - they will both be deserialised properly

[aslakhellesoy]

Please hold off on merging this while I fix the failing tests. I'm also adding a bugfixes for enum fields and JSON.

[aslakhellesoy]

With this change, bytes values have to be encoded as Encoding::ASCII_8BIT, which in Ruby means "binary" encoding. Any other encoding will cause the value to be Base64 decoded.

This is because the JSON representation of bytes is always Base64. When deserialising JSON we don't have the protobuf field info, so the decoding is (unfortunately) happening inside Protobuf::Field::BytesField, based on whether the value is a "binary" string or a "base64" string.

[film42]

The underscore thing seems like we'd want to make that optional?

[film42]

Would it make sense to make this configurable at the from_json method? You can specify if your uses Base64 encoded binary blobs? Maybe true by default but allow the caller to turn it off? That way we don't have to guess.

[jackorp]

Ping... Hi, what's the status on this? I am packaging this library for Fedora as a part of an effort to update cucumber and it would be nice if this PR and PR#415 were merged.