[FEATURE] Add dynamic variable substitution for Normalization processor parameters

[martin-gaievski]

Is your feature request related to a problem?

The Normalization processor currently only accepts static parameters during pipeline creation. This limitation becomes problematic when parameter values need to be determined dynamically during search execution.

Proposed solution is to enable dynamic parameter substitution in the Normalization processor by:

• allowing placeholder variables in pipeline definitions
• supporting runtime value substitution during query execution

For example pipeline definition with normalization processor can look like this:

{
    "description": "Post processor for hybrid search",
    "phase_results_processors": [
        {
            "normalization-processor": {
                "normalization": {
                    "technique": "min_max"
                },
                "combination": {
                    "technique": "arithmetic_mean",
                    "parameters": {
                        "weights": [
                             ${keyword_weight},
                             ${neural_weight}
                         ]
                    }
                }
            }
        }
    ]
}

when one of the processors (e.g. ML Inference processor) set values in the pipeline context, the Normalization processor would substitute these variables with actual values.

What alternatives have you considered?

External parameter management: While possible to handle parameter values outside OpenSearch, this approach adds maintenance overhead and complexity to the system.

Do you have any additional context?

This is one of the building blocks for hybrid search optimizer that has been requested in opensearch-project/OpenSearch#16775

[wrigleyDan]

Great, that's very useful to overcome the "staticness" of hybrid search and open the door to dynamic query-based settings that are determined outside of the normalization process 👍