[DOCS] Clarifies description of num_top_feature_importance_values

[lcawl]

This PR clarifies the descriptions for analysis.classification.num_top_feature_importance_values and analysis.regression.num_top_feature_importance_values in the create data frame analytics job API.

Preview: http://elasticsearch_52246.docs-preview.app.elstc.co/guide/en/elasticsearch/reference/master/put-dfanalytics.html

[elasticmachine]

Pinging @elastic/ml-core (:ml)

[elasticmachine]

Pinging @elastic/es-docs (>docs)

[valeriy42]

I noticed that you removed the reference to the SHAP algorithm. I think it should be described somewhere in the documentation.

Furthermore, we should say that we may return less feature importance values than the parameter specifies if there are no more features available in the model or if the feature importance of other features for this document is zero.

[lcawl]

I noticed that you removed the reference to the SHAP algorithm. I think it should be described somewhere in the documentation.

It still exists in the pages that are linked from these API property descriptions (e.g. http://elasticsearch_52246.docs-preview.app.elstc.co/guide/en/machine-learning/master/dfa-classification.html#dfa-classification-feature-importance).

[lcawl]

Furthermore, we should say that we may return less feature importance values than the parameter specifies if there are no more features available in the model or if the feature importance of other features for this document is zero.

IMO That seems like the type of information that will require further explanation and should be added in the https://www.elastic.co/guide/en/machine-learning/current/dfa-regression.html#dfa-regression-feature-importance and https://www.elastic.co/guide/en/machine-learning/current/dfa-classification.html#dfa-classification-feature-importance pages. I've created elastic/stack-docs#881