LTR feature logger

[afoucret]

The PR provides tools for LTR query feature extraction into Eland.

1. Some base class have been defined to declare the LTR config:

ltr_config = LTRModelConfig(
  feature_extractors=[
    QueryFeatureExtractor(
      feature_name="title_bm25", 
      query={"match": {"title": "{{query}}"}},
    ),
    QueryFeatureExtractor(
      feature_name="popularity",
      query={
        "script_score": {
          "query": {"exists": {"field": "popularity"}},
          "script": {"source": "return doc['popularity'].value"},
        }
      },
    )
  ]
)

2. A feature logger can be created using:

feature_logger = FeatureLogger(es_client="http://localhost:9200",  es_index="my-index", ltr_model_config=ltr_model_config)

3. Document features can be extracted for a set of query params using:

feature_logger.extract_features(
  query_params={ "query": "my search query" },
  doc_ids= ["doc-1", "doc-2"] 
)