[R-package] Added instructions for generating R test coverage

[jameslamb]

Really excited to say that I've discovered that covr (the most popular library for generating test coverage for R projects) now works well with lightgbm's R package! I had created jameslamb/lightgbm-r-coverage to work around the previous issues I had and I am VERY excited to archive that repo.

In this PR, I propose adding instructions to the R package's README explaining how contributors can generate coverage reports. I think this will be a valuable way for us to catch edge cases that are not covered by our current tests, and all the other good things you can do with code coverage.

At the time of creating this PR, the R package is 64.48% covered.

lightgbm Coverage: 64.48%
R/lgb.plot.importance.R: 0.00%
R/lgb.prepare_rules.R: 0.00%
R/lgb.prepare_rules2.R: 0.00%
R/lgb.prepare.R: 0.00%
R/lgb.prepare2.R: 0.00%
R/lgb.unloader.R: 0.00%
R/readRDS.lgb.Booster.R: 0.00%
R/saveRDS.lgb.Booster.R: 0.00%
R/lgb.cv.R: 57.14%
R/lgb.Predictor.R: 57.66%
R/lgb.Booster.R: 62.58%
R/callback.R: 67.48%
R/utils.R: 70.63%
R/lgb.Dataset.R: 72.02%
R/lgb.train.R: 75.94%
R/lightgbm.R: 95.45%
R/aliases.R: 100.00%
R/lgb.importance.R: 100.00%
R/lgb.interprete.R: 100.00%
R/lgb.model.dt.tree.R: 100.00%
R/lgb.plot.interpretation.R: 100.00%

I've attached a detailed coverage report for those who are curious (GitHub will not let you attach a .html file, so had to zip it).

coverage.html.zip

[StrikerRUS]

LGTM! Thanks!

[jameslamb]

Just rebased to current master. I added two more files to .gitignore that were missed before.

@StrikerRUS normally I wouldn't merge an R-package PR with only your approval but I think for this once it's ok. Do you agree?

[StrikerRUS]

@jameslamb

normally I wouldn't merge an R-package PR with only your approval but I think for this once it's ok. Do you agree?

Sure!