Test if optimizer returns a best guess as expected #14
Description
Naturally, the `step` function prints every model it's tested into the console, verbose both suppresses that and warnings on poorness of fit.
I'm unsure if a NULL set is possible, as a formal error would generate NAs later on, and a non-error warning would imply there wasn't convergence, or that there were non-integer successes, but there may be a 'best guess' by the optimizer which would also not generate a NULL set.
Originally posted by @ryan-odea in #12 (comment)