[R-package] Add the ability to predict on lgb.Dataset in Predictor$predict()

[jameslamb]

Summary

Currently, the data argument of Predictor$predict() accepts a matrix, dgCMatrix, or filepath to a dataset stored in a file. Should add the support to pass in a lgb.Dataset object.

Motivation

Other parts of the lightgbm R package require or strongly encourage the use of an lgb.Dataset. If users already have code to convert their data into a lgb.Dataset, they should be able to use that same preprocessing to pass data to predict() to generate predictions.

Description

To close this issue, add support for lgb.Dataset in the Predictor class (R-package/R/lgb.Predictor.R) and add corresponding tests confirming that the predictions that are returned are correct.

[jameslamb]

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 actively working on implementing this feature.