The previous version of the spec adds fractional digit-related option…

…s to _nfOpts_ whenever there are fractional-styled subsecond units, even if the value of all of those units is 0. In practice, this means that whenever numeric seconds are formatted, the _nfOpts_ for maximum and minimum fractional digits get passed to the NumberFormat. In order to avoid normative changes, I've removed everything that tests for whether non-zero fractional-styled units exist.