Feature/minimal mlflow inference examle

[tibrk]

This pull request introduces a minimal example for deploying and testing an MLflow-tracked model as a KServe InferenceService using both v1 and v2 inference protocols.

Deployment and Protocol Support:

• Added example InferenceService manifests for both v1 and v2 KServe inference protocols, demonstrating how to deploy an MLflow model using KServe, including placeholders for customization. [1] [2] [3]
• Added a deep dive into both protocol versions. Explaining differences and providing a notebook to get some hands on experience with both.

Testing and Usage:

• Added a Python script (test_inference_service.py) to send authenticated inference requests to the deployed service, supporting both v1 and v2 protocols and reading request bodies from JSON files.

Sample Request Bodies:

• Provided sample JSON request bodies formatted for both v1 and v2 inference protocols, using realistic data from the mobile price classification dataset. [1] [2]

Documentation:

• Added a comprehensive README.md explaining prerequisites, deployment steps, environment variable configuration, testing instructions, and the request body format for the minimal MLflow model inference example.

[tmvfb]

@tibrk, thanks for the contribution, testing this! Before testing I have a few nitpicks/suggestions:

1. Could we please rename the files to look conventionally (.ipynb should be snake case, .yaml should be kebab case, we could also shorten InferenceService to isvc).

2. If we can remove the duplicated v2-mlflow-inference-body.json, let's go for it.

3. I'd suggest clearer names for folders, e.g.

   mlflow-kserve-minimal-example
   mlflow-kserve-inference-protocols
   

4. Maybe we'll need a short README in the /serving folder after this PR, since we'll have 4 subfolders already.