RemoveMissingData class isn't applying training information to test data

[jchristo12]

When fitting a RemoveMissingData class, the columns that are removed due to too much missing data are not the same columns that are removed from new data run thru the class.

In other words, the training data should find the columns that should be removed and these same columns are the ones that should be dropped from the test data.

Instead, the class is finding the columns in the test data with too much missing data and removing those.

[jchristo12]

fix may be in place but is outside the pipeline; still investigating if this can be handled within the pipeline

[jchristo12]

Removing bug tag since a workaround is in place.