Adds decision tree classes which can be derived from trained forests.
Changes
Add pytest-xdist as a dependency and a command xtest for running distributed tests
Move bindings from ensemble module to top level module, since they are now shared by ensemble and tree submodules
Add estimators_ property to forests which return a list of tree estimators
Add a tree submodule
Add a BaseGRFTree with common tree traversing methods, matching the sklearn BaseDecisionTree methods
Add GRFTree classes for relevant forest implementations. These trees have classmethods for instantiating from trained forests. The classes use the same serialized forest structure so that we can reuse the grf bindings by calling into a forest of just one tree.
Rename fixtures -> _fixtures
Make tests easier to copy/paste by renaming instances to forest and tree
Fix a bug on local linear forest where ll_split_variables_ was not being set properly
Adds decision tree classes which can be derived from trained forests.
Changes
xtestfor running distributed testsestimators_property to forests which return a list of tree estimatorstreesubmodulell_split_variables_was not being set properly