I haven't explained the reason of the error estimation in
Range#step for Float;

double n = (end - beg)/unit;
double err = (fabs(beg) + fabs(end) + fabs(end-beg)) / fabs(unit) * epsilon;

The reason is as follows. (including unicode characters)
This is based on the theory of the error propagation;
http://en.wikipedia.org/wiki/Propagation_of_uncertainty

If f(x,y,z) is given as a function of x, y, z,
¶§f (the error of f) can be estimated as:

¶§f^2 = |ĘŖf/ĘŖx|^2*¶§x^2 + |ĘŖf/ĘŖy|^2*¶§y^2 + |ĘŖf/ĘŖz|^2*¶§z^2

This is a kind of `statistical' error.  Instead, `maximum' error
can be expressed as:

¶§f = |ĘŖf/ĘŖx|*¶§x + |ĘŖf/ĘŖy|*¶§y + |ĘŖf/ĘŖz|*¶§z

I considered the latter is enough for this case.
Now, the target function here is:

n = f(e,b,u) = (e-b)/u

The partial differentiations of f are:

ĘŖf/ĘŖe = 1/u
ĘŖf/ĘŖb = -1/u
ĘŖf/ĘŖu = -(e-b)/u^2

The errors of floating point values are estimated as:

¶§e = |e|*¶Ň
¶§b = |b|*¶Ň
¶§u = |u|*¶Ň

Finally, the error is derived as:

¶§n = |ĘŖn/ĘŖe|*¶§e + |ĘŖn/ĘŖb|*¶§b + |ĘŖn/ĘŖu|*¶§u
= |1/u|*|e|*¶Ň + |1/u|*|b|*¶Ň + |(e-b)/u^2|*|u|*¶Ň
= (|e| + |b| + |e-b|)/|u|*¶Ň

Masahiro Tanaka