[Learn to Rank] LTR explain should use shapley feature importance values if possible

[benwtrent]

Description

XGBoost and other types of models can actually provide information indicating the individual branch and feature importance in the tree leaves.

Inference in Elasticsearch can utilize this information to calculate shapley importance of each feature during inference. This should be done if possible as it provides valuable insights into what features were actually important for the given prediction, not just what features were used.

[elasticsearchmachine]

Pinging @elastic/ml-core (Team:ML)

[elasticsearchmachine]

Pinging @elastic/es-search-relevance (Team:Search Relevance)