na_option is not used in DataFrame.rank()

[lbeltrame]

Currently na_option is unused in rank(). It would be nice to have at last "top" or "bottom" along with "keep" (it's actually useful as some published algorithms need this feature).