Boost Review: code duplication part 2.

[jzmaddock]

Code duplication: Importantly, I don't understand why the trait bias_v is even needed, given that it duplicates the value stored in std::numeric_limits<...>::digits/digits10/maxdigits10. Same story with the emin_v and emax_v traits compared to the std::numeric_limits<...>::min_exponent/min_exponent10 and std::numeric_limits<...>::max_exponent/max_exponent10. This seems like some code duplication that would be prone to causing confusion later. Why do multiple definitions of traits exist that have the same values? There may be other cases of this as well. I would love to see either std::numeric_limits versions being defined in terms of the other traits where possible, or vice-versa. I find it confusing to have 2 bespoke definitions.