Support all NeptuneML API command parameters in neptune_ml magics, accept unified JSON blob for parameter input

[michaelnchin]

Issue #, if available: #65, #103, #187

Description of changes:

• Added support for all of the NeptuneML API command parameters listed in the Neptune documentation.
• Support unified JSON object as a single variable to supply parameter inputs for each phase (excluding Export)

JSON parameter blob usage:

1. Define the JSON blob as a notebook variable, following the format below. Note that this example defines the bare minimum set of parameters required for each step; you may define any number of additional optional parameters listed at the NeptuneML API.

training_job_name=neptune_ml.get_training_job_name('link-prediction')  # example job ID for substitution 

ml_params = {
    "dataprocessing": {
        "id": f'{training_job_name}',
        "configFileName": "training-data-configuration.json",
        "inputDataS3Location": f"{export_results['outputS3Uri']}",  # ensure that the export job has finished first, or this will fail
        "processedDataS3Location" : f"{str(s3_bucket_uri)}/preloading"
    },
    "training": {
        "id": f'{training_job_name}',
        "dataProcessingJobId": f'{training_job_name}',
        "trainingInstanceType": "ml.p3.2xlarge",
        "trainModelS3Location": f"{str(s3_bucket_uri)}/training",
        "maxHPONumberOfTrainingJobs": 2,
        "maxHPOParallelTrainingJobs": 2
    },
    "endpoint": {
        "id": f'{training_job_name}',
        "mlModelTrainingJobId": f'{training_job_name}'
    },
    "modeltransform": {
        "id": f'{training_job_name}',
        "mlModelTrainingJobId": f'{training_job_name}',
        "dataProcessingJobId": f'{training_job_name}',
        "modelTransformOutputS3Location": f"{str(s3_bucket_uri)}/modeltransform"
    }
}

2. Pass the JSON variable into the cell body of the desired %%neptune_ml cell magic command, using the cell variable injection syntax. Re-use with as many steps as needed.

%%neptune_ml dataprocessing start --wait --store-to processing_results 
${ml_params}

New %neptune_ml Arguments:

%neptune_ml dataprocessing:

• --prev-job-id - The job ID of a completed data processing job run on an earlier version of the data.
• --instance-type - The type of ML instance used during data processing.
• --instance-volume-size-in-gb - The disk volume size of the processing instance.
• --timeout-in-seconds - Timeout in seconds for the data processing job.
• --model-type - Heterogeneous graph model (heterogeneous) or knowledge graph model (kge).

%neptune_ml training:

• (REQUIRED) --max-hpo-number - Maximum total number of training jobs to start for the hyperparameter tuning job.
• (REQUIRED) --max-hpo-parallel - Maximum number of parallel training jobs to start for the hyperparameter tuning job.
  ---prev-job-id - The job ID of a completed model-training job that you want to update incrementally based on updated data.
• -model_name - The model type for training. If not specified, the model-training job will use the same modelType used in the data processing step.
• -base-processing-instance-type - The type of ML instance used in preparing and managing training of ML models.
• --instance-volume-size-in-gb - The disk volume size of the training instance.
• --timeout-in-seconds - Timeout in seconds for the training job.

%neptune_ml modeltransform:

• --job-id - A unique identifier for the new job.
• --s3-output-uri - The URI of the S3 bucket/location to store your transform result.
• --data-processing-job-id - The job Id of a completed data-processing job.
• --model-training-job-id - The job Id of a completed model-training job.
• --training-job-name - The name of a completed SageMaker training job.
• --base-processing-instance-type - The type of ML instance used in preparing and managing training of ML models.
• --base-processing-instance-volume-size-in-gb - The disk volume size of the new training instance.

Note that you must now specify either:

a) --data-processing-job-id AND --model-training-job-id
b) --training-job-name

%neptune_ml endpoint:

• --model-training-job-id - The job Id of a completed model-training job.
• --model-transform-job-id - The job Id of a completed model-transform job.
• --update - Indicates that this is an update request.
• --model-name - Model type that was used for training.
• --instance-type - The type of ML instance used.
• --instance-count - The minimum number of Amazon EC2 instances to deploy to an endpoint for prediction.
• --neptune-iam-role-arn, -volume-encryption-kms-key - See "Shared Security Parameters" below

Note that you must now specify either:

a) --model-training-job-id
b) --model-transform-job-id

Shared security parameters for dataprocessing, training, and modeltransform:

• --sagemaker-iam-role-arn - The ARN of an IAM role for SageMaker execution.
• --neptune-iam-role-arn - The ARN of an IAM role that provides Neptune access to SageMaker and Amazon S3 resources.
• --subnets - The IDs of the subnets in the Neptune VPC.
• --security-group-ids - The VPC security group IDs.
• --volume-encryption-kms-key - The Key Management Service(KMS) key used by SageMaker to encrypt data on the storage volume attached to the ML compute instances that run the job.
• --s3-output-encryption-kms-key - The KMS key that SageMaker uses to encrypt the output of the job.

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.