Use a typedef for the elements of floating_point_components

[eyalroz]

We currently hard-code the use of int_fast64_t in floating_point_components, and also cast to that hard-coded type in several places. Let's use a type definition istead (also in light of the suggestions in PR #221) .