Honest trees not compatible with BEST_FIRST_GLOBAL

[PSSF23]

I encountered the following error:

Traceback (most recent call last):
  File "/home/hao/ydf/real_honest_ydf.py", line 169, in <module>
    model = learner.train(X_t)
  File "/home/hao/miniconda3/envs/skt/lib/python3.9/site-packages/ydf/learner/specialized_learners.py", line 2947, in train
    return super().train(ds=ds, valid=valid, verbose=verbose)
  File "/home/hao/miniconda3/envs/skt/lib/python3.9/site-packages/ydf/learner/generic_learner.py", line 345, in train
    return self._train_imp(ds, valid, verbose)
  File "/home/hao/miniconda3/envs/skt/lib/python3.9/site-packages/ydf/learner/generic_learner.py", line 501, in _train_imp
    return self._train_from_dataset(ds, valid)
  File "/home/hao/miniconda3/envs/skt/lib/python3.9/site-packages/ydf/learner/generic_learner.py", line 582, in _train_from_dataset
    cc_model = learner.Train(**train_args)
ValueError: INVALID_ARGUMENT: honest trees are not (yet) supported with growing_strategy_best_first_global strategy.

It seems that honesty is not compatible with growing_strategy="BEST_FIRST_GLOBAL"? I don't understand this limitation, as honesty doesn't have anything to do with how the nodes are split.

[rstz]

Hi, there is no technical limitation for this - I'll look into how much work it will be to add this.

BTW: Make sure you take a look at d5c4642 as it fixes an issue with our honest implementation.