[dask] Support init_model

[StrikerRUS]

Summary

LightGBM allows you to provide initial model for continued training.

This option should be supported in the Dask interface.

Motivation

This change would bring the Dask interface closer to full feature parity with the non-Dask interface, so that users who'd otherwise like to use Dask don't have to avoid it because init_model is missing.

References

LightGBM/python-package/lightgbm/sklearn.py

Lines 252 to 253 in e5c3f7e

	init_model : string, Booster, LGBMModel or None, optional (default=None)
	Filename of LightGBM model, Booster instance or LGBMModel instance used for continue training.

LightGBM/python-package/lightgbm/engine.py

Lines 159 to 164 in d6ebd06

	if isinstance(init_model, str):
	predictor = _InnerPredictor(model_file=init_model, pred_parameter=params)
	elif isinstance(init_model, Booster):
	predictor = init_model._to_predictor(dict(init_model.params, **params))
	else:
	predictor = None

LightGBM/python-package/lightgbm/dask.py

Line 858 in 296397d

# DaskLGBMRegressor support for callbacks and init_model is not tested

[StrikerRUS]

Closed in favor of being in #2302. We decided to keep all feature requests in one place.

Welcome to contribute this feature! Please re-open this issue (or post a comment if you are not a topic starter) if you are actively working on implementing this feature.