Relax kadet output type requirement.

[Jean-Daniel]

Proposed Changes

There is 2 issues with requiring BaseObj in kadet.

BaseObj as return type for main()

kadet Input expects just a mapping of filename -> dict. Requiring a BaseObj is confusing as it implies kadet expects a single object and not a list of files. This change adds support for raw dict as return type. It simplifies implementation of kadet modules as it remove the need to build a BaseObj just to returns a mapping of filenames to BaseObj.

For instance, the nginx example may be rewrite as:

def main():
    return {
      "nginx_deployment": kubelib.Deployment(name=name, labels=labels, containers=[nginx_container]),
      "nginx_service": kubelib.Service(name=name, labels=labels, ports=[svc_port], selector=svc_selector)
    }

instead of

def main():
    output = kadet.BaseObj()
    output.root.nginx_deployment = kubelib.Deployment(name=name, labels=labels, containers=[nginx_container])
    output.root.nginx_service = kubelib.Service(name=name, labels=labels, ports=[svc_port], selector=svc_selector)
    return output

BaseObj as a required type for generated objects.

While using BaseObj may be convenient to build a complex nested dict, this is not the single way to generate such structure, and imposing it prevents exploring other frameworks. This change also relax this restriction and support simple dict in addition of BaseObj so any generator/kadet module would be able to generates dict as they see fit.

[ramaro]

Very good point and thanks again @Jean-Daniel