Closes #211 -- handle missing input consistently

[MichaelChirico]

This allows the following syntax to work without error:

DF = data.frame(json = c('{"hey": 1, "you": 2}', NA_character_, '{"hey": 3, "you": 4}', stringsAsFactors = FALSE)
do.call(rbind, lapply(DF$json, fromJSON))
#    hey you
# 2    1   2
# 21   3   4