higher_is_better for InfoLM should change depending on information_measure

[grahamannett]

Not sure if there are other metrics this will apply to but for InfoLM, if the information_measure is l2, then the distance will always positive, but infolm.higher_is_better shows True. I am pretty sure that is no longer the case compared to the base measure which is KL in which case trending towards 0 from vals < 0 is improvement

[Borda]

That is a good point and should be fixed. Are you interested in sending a PR?
cc: @lucadiliello

[grahamannett]

Sure, will try and send a PR in the next few days.

Was thinking/wondering if its an issue for other metrics but not sure if I will have time to go through and look into others/understand if it applies in other cases. This one specifically made some metrics for runs very confusing!