Flag potential improper use of by variables

[mstackhouse]

Description

When by variables are provided into Tplyr, there's a complete run to fill NA values. There's an inherent assumption we've made where you won't provide two by variables that are 1:1. For example, VISITNUM and VISIT. This is because Tplyr can autoamtically detect that AVISITN should be using for sorting AVISIT, based on ADaM assumptions.

If you do provide VISITNUM and VISIT, it will duplicate all of those records and essentially cartesian join the results because of the dplyr::complete() calls we run. This is necessary to an extent, because we want to provide the 0 rows of factor combinations if with no results in the data, but makes these scenarios a bit unintuitive and confusing.

As a preventative measure - we should introduce a warning if we notice a large proportional increate of records due to the complete. Gauging what that ratio increase is a little tough - but maybe somehting like if we notice that rows increase by 50% then produce a warning.

[mstackhouse]

Closed via #174