PySparkProcessor - Possibility to choose different instance types for the driver node and the worker nodes

[HarryPommier]

Describe the feature you'd like
It would be nice to have the possibility to choose different instance types for the driver node and the worker nodes when using the PySparkProcessor.

How would this feature be used? Please describe.

pyspark_processor = PySparkProcessor(
    base_job_name=...,
    framework_version=...,
    role=...,
    driver_instance_type="ml.m5.4xlarge",
    worker_instance_type="ml.m5.large",
    instance_count=...,
    sagemaker_session=pipeline_session,
    max_runtime_in_seconds=...,
)

Describe alternatives you've considered
It is possible to choose a high-memory instance for all instances but it could be unnecessarily costly for the user.

Additional context
Some pyspark operations (e.g. .toPandas()) are memory expensive for the driver node.

[j-adamczyk]

+1, this is quite weird not being able to set this