internal attributes dont perfectly match what sklearn expects

[falkben]

for instance:

self.estimators_ contains a list of trees in sklearn

However, we store multiple trees in each bin, with multiple bins representing the forest. also, trees are interleaved within the bins.

At this point an time, we are pretty fundamentally incompatible with this internal attribute (at least for our binned implementations). There may still exist some solution where we provide an interface to get individual trees out of the bins...?