Execution inputs as container arguments for processing jobs

I'm trying to use execution inputs as container arguements for my processing job:

```
execution_input = ExecutionInput(
schema={
"IngestaJobName": str,
"PreprocessingJobName": str,
"InferenceJobName": str,
"Fecha": str,
}
)
```

```
#Call step
ingesta_step = ProcessingStep(
inference_config["ingesta_step_name"],
processor=ingesta_processor,
job_name=execution_input['IngestaJobName'],
inputs=inputs_ingesta,
outputs=outputs_ingesta,
container_arguments=["--fecha", "$$.Execution.Input['Fecha']"],
container_entrypoint=["python3", "/opt/ml/processing/input/code/"+inference_config["ingesta_function"]],
)
```

I've also tried to replace container_arguments for `["--fecha", execution_input["Fecha"]] `

But in both cases it doesn't work.

### Use Case

When I lunch a new execution of my state machine, it would be useful to get some execution inputs as a container argument in order to define some parameters of intereset that will be define the behaviour of the step directly by the execution input without updating the state machine definition

---

This is a :rocket: Feature Request

Start by tracing the ProcessingStep container_arguments entry point and how ExecutionInput values are represented in the SDK. Reproduce the supplied processing-job example, then inspect the related implementation and tests if available. Done means an execution input can reach the processing container as an argument without changing the state machine definition.

Vom Indexierungsmodell aus dem Issue-Text verfasst.