Make @interval return a Float64 interval

[dpsanders]

Since Float64 is the most common case, I suggest that @interval do what @floatinterval does now.

[lbenet]

I think it already does that by default. Consider the following:

julia> using IntervalArithmetic

julia> a = big"0.1"
1.000000000000000000000000000000000000000000000000000000000000000000000000000002e-01

julia> f(x) = @interval a + x
f (generic function with 1 method)

julia> g(x) = @biginterval a + x
g (generic function with 1 method)

julia> @format full
6

julia> f(0.0)
Interval(0.09999999999999999, 0.1)

julia> g(0.0)
Interval(9.999999999999999999999999999999999999999999999999999999999999999999999999999914e-02, 1.000000000000000000000000000000000000000000000000000000000000000000000000000002e-01)

Or do you mean something else?

[dpsanders]

julia> @macroexpand @interval 0.1
:((IntervalArithmetic.Interval)((IntervalArithmetic.atomic)((IntervalArithmetic.Interval){IntervalArithmetic.parameters.precision_type}, 0.1)))

julia> @macroexpand @floatinterval 0.1
:((IntervalArithmetic.Interval)((IntervalArithmetic.atomic)((IntervalArithmetic.Interval){Float64}, 0.1)))

@interval uses the nasty global precision business -- probably we should just get rid of that?
That leads to a factor 4 slowdown in

g(x) = @interval 0.1 + x
versus
h(x) = @floatinterval 0.1 + x