Swap out entropy for other metrics to determine splits for tree based models

[arunaryasomayajula]

Is your feature request related to a problem? Please describe.

Request/Question 1: Swap out entropy for other metrics to determine splits

What we want to do: Within the tree family of estimators, we want to change the splitting metric. While entropy is great for splits since it often finds the most separation, we would like to create the gradient as usual to continue fitting downstream trees but use different buckets for calculating the gradient.

Why: There are cases where we aim to optimize a certain segment of the population, so we do not care about the best separation across all buckets. We want this to be reflected in the variable bucketing (tree splits). Therefore, we are interested in swapping out the metric for determining the best splits to something other than entropy.

https://support.h2o.ai/a/tickets/109575

Describe the solution you'd like
A clear and concise description of what you want to happen.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered, if applicable.

Additional context
Add any other context or screenshots about the feature request here. If there's a reference (paper, book, etc) for this feature, please add that here.

H2O.ai Devs only
If there is a support ticket associated with this issue, please post the link here.