matchfailed() tests, documentation

[benthestatistician]

I experienced an issue where matchfailed() returned NA for a small number of observations, breaking a script. (Unfortunately I couldn't stop to investigate.) Hypothesis: this is how it records an exact matching category with no treatments or controls, such that it wouldn't make sense to try to make matches within it.

• Whether that's correct or not should be recorded in the Roxygen block describing the function, and there should be tests to confirm the behavior.
• Accompanying that should be a few simple tests to confirm thatmatchfailed() does indeed mark subclasses in which matching failed, and doesn't mark those in which it succeeded.
• The person who does this should also test any other scenarios occurring to them where the function might unexpectedly return an NA as opposed to a TRUE or FALSE.