* Fix tensorflow MNIST dataset download bug in multiple trials * Modify the maxTrialNum and trialConcurrency in config
[Kubeflow training service] Use Kubernete API server to replace kubectl dependency
* Remove unused kubernetesServer config entry in config file and schema validation
* Add distributed mnist training example, to show how to perform distributed training on kubeflow for NNI * rename folder name to mnist_distributed * Remove duplicated is_chief check