[Feature request] Amend ClusterPurpose class

ClusterPurpose class currently has -

* `FAST_PROD` will provision azureml-fe instances to reverse proxy traffic, 3 nodes
* `DEV_TEST` will provision the model without azureml-fe, on a single node

I've worked with a few customers that would be much happier with a middleground option, one that would deploy the model on multiple nodes but without fronting it with azureml-fe. They may have their own ingress setup or they may want to expose a naked service (which would be fronted by API Management for example). We could call this `PROD` or `STANDARD_PROD`, because `NAKED_PROD` may sound a little strange even if it describes the use case perfectly.

This new deployment model would also allow folks to benchmark their models in a realistic way (true requests/sec figure).

---
#### Document Details

⚠ *Do not edit this section. It is required for docs.microsoft.com ➟ GitHub issue linking.*

* ID: 0b62a589-d158-cce2-5167-2ef641c39b3d
* Version Independent ID: 3ba04a54-fe40-8f55-3981-b7ad5a3c845a
* Content: [azureml.core.compute.AksCompute.ClusterPurpose class - Azure Machine Learning Python](https://docs.microsoft.com/en-us/python/api/azureml-core/azureml.core.compute.akscompute.clusterpurpose?view=azure-ml-py)
* Content Source: [AzureML-Docset/stable/docs-ref-autogen/azureml-core/azureml.core.compute.AksCompute.ClusterPurpose.yml](https://github.com/MicrosoftDocs/MachineLearning-Python-pr/blob/live/AzureML-Docset/stable/docs-ref-autogen/azureml-core/azureml.core.compute.AksCompute.ClusterPurpose.yml)
* Service: **machine-learning**
* Sub-service: **core**
* GitHub Login: @DebFro
* Microsoft Alias: **debfro**

Beginne mit der Dokumentation der Klasse ClusterPurpose und der verknüpften Quelle AzureML-Docset/stable/docs-ref-autogen/azureml-core/azureml.core.compute.akscompute.clusterpurpose.yml. Verfolge die Implementierung des Python SDK und die Bereitstellungspfade, um zu bestimmen, wie ein neuer Zweck eine Bereitstellung mit mehreren Knoten ohne azureml-fe darstellen sollte. Erledigt ist die Aufgabe, wenn Verhalten und Name abgestimmt sind, das Bereitstellungsmodell funktioniert und die Dokumentation der Klasse aktualisiert wurde.

Vom Indexierungsmodell aus dem Issue-Text verfasst.