add alternative feature importance metric for logistic regression

Triage seems to use odds ratios as feature importance measures for logit models. This can be confusing when these are naively compared with importances from other models, which typically use 0 rather than 1 as the baseline/ no effect. Maybe absolute (standardized) coefficients or t-values make sense here.

The issue names no files, tests, or entry points. First locate the logistic-regression feature-importance implementation and its tests, then clarify whether the project should support absolute standardized coefficients, t-values, or another metric and how the result should be compared with other model importances; done means an agreed metric is implemented and covered by tests.

Written by the indexing model from the issue text.