approx_eq_ulps() allows specification of epsilon as an integer number
of Units in the Last Place (ULPs) difference between the two floating
point values
default approx_ulps() is set to 8.
approx_eq() function continues to use epsilon method, although I
recommend further commits and a migration towards the ULPs method.
These macros yield readable error messages as test assertions for ApproxEq
types. They can be invoked as:
assert_approx_eq!(a, b);
assert_approx_eq_eps!(a, b, eps);
Fixes#40.
