Decode decimal types in one shot

[mborland]

For example in frexp10 we have this:

auto num_exp {num.biased_exponent()};
auto num_sig {num.full_significand()};
detail::normalize<T>(num_sig, num_exp);

This involves decoding the combination field twice. We should be able to combine these into a single operation called to_components that returns the already existing structure of components.