The goal is to guess whether codes are short_code or
decimal_code form. Currently condense works, but not with the
icd look-up table currently in use. Of note, validation is a bit
different here, since we don't know the type until after we guess. We could
look for where both short_code and long are invalid, and otherwise assume
valid, even if the bulk are short_code. However, it may be more useful to
check validity after the guess.
guess_short(x_, short_code = NULL, n = 1000L, icd_name = NULL)
single logical value, TRUE if input data are predominantly
short_code type. If there is some uncertainty, then return
NA.
Very quick heuristic, ploughs through ten million codes in less than one second and will stop more quickly if it finds a '.'.