[META] ML Inference Processor Flexibility Enhancements II

[mingshl]

Is your feature request related to a problem?

• Currently, search response processor can add prediction results to new fields to extensions. Try to enable option to save the model prediction to the extension part in the search response.
• For rerank use cases, we need to pass on the query_text to the model input dataset along with the documents in the search response. Try to enable reading value from query body in model config.
• For rerank use cases, we also need a rescore processor that can rank the search hit and update the new scores.

All META Issues for ML Inference Processor:
#2839

What solution would you like?
A clear and concise description of what you want to happen.

What alternatives have you considered?
A clear and concise description of any alternative solutions or features you've considered.

Do you have any additional context?
Add any other context or screenshots about the feature request here.

[andrross]

[Catch All Triage - 1, 2, 3]

[mingshl]

closing this issue as features are implemented in 2.18