`fit_aoa` features

[mcfrank]

fit_aoa fails when there are NAs - so if you fit everything it will fail.

eng_ws_data <- get_instrument_data(language = "English (American)",
                                   form = "WS",
                                   item_info = TRUE,
                                   administration_info = TRUE)

eng_ws_aoa_all <- fit_aoa(eng_ws_data) # chokes
eng_ws_aoa <- fit_aoa(filter(eng_ws_data, item_kind == "word"))

also - do we want the bayesian method from the book?