Create a pipeline schema

[isra17]

Right now we parse each individual parameters, check if it's a dataclass and then create a schema for it. It would be simpler and more robust if we created a schema for all the pipeline parameters and loaded it with the message arguments at once. This would avoid workaround to support extra type at the parameter level (See https://github.com/Flared/saturn/pull/139/files).

ie:

@classmethod
def instancify_args(args: dict[str, object], pipeline: Callable) -> dict[str, object]:
  pipeline_schema = cls.schema_for_pipeline(pipeline)
  return dataclasses.asdict(pipeline_schema.load(args))

This way we let desert deal with optionals, dict, etc.