v0.9.0

Breaking Changes

• InputObject no longer has a serialize method - this is now handled by a
  SerializableArgument impl instead, which is generated by the InputObject
  derive.
• Query has been renamed to Operation to make it clear it's used for both
  queries & mutations.
• Query::new is now Operation::query

New Features

• InputObjects can now be derived and will be generated by querygen.
• Querygen output is now tested more thoroughly - should be less changes
  required by users just to get it to compile.
• Cynic now supports running & generating code for mutations.

Removed Features

• Removed the optimised_query_modules feature from codegen, as it invovled
  more code than it was worth to keep it around. Functionally this should make
  no difference, though it may change performance characteristics of compiling
  cynic code. Didn't seem to make a significant difference when I was using it
  though.

Bug Fixes

• Fixed a compile issue in the generated query_dsl for schemas with fields
  with > 1 required argument.
• Fixed an issue that required users to add serde_json to their dependencies.
  We now re-export it as cynic::serde_json and use that in our derive output.
• querygen now adds rename_all="SCREAMING_SNAKE_CASE" to Enums by default -
  the GQL convention is to have them in this format and querygen was already
  doing the transformation into the PascalCase rust usually uses so this
  should make things more likely to work by default.
• Removed fontawesome from the querygen HTML. Think I added this along with
  bulma but it's not being used, and adds 400kb to the payload.
• Fixed a bug where querygen would not snake case field names when generating
  QueryFragments.
• querygen will now take references to arguments rather than ownership (which
  didn't work for most non-enum types).
• Fixed an issue where querygen was adding ID literals as Strings in arguments,
  rather than IDs.