As discussed, here are @MechCoder s changes from PR 2493 played onto master, with necessary adjustments made (sample weights weren't present at the time).
This PR is WIP because I haven't added the doc changes yet. Otherwise everything is done.
replayed manojs changes in pr2493 onto master
Thanks @eickenberg :-)
OK, I just realized that metrics has been refactored completely, so this will take a few more moments ....
Closing this, I made a better one just now.