Approximate equality with intervals

[mforets]

in the comparison below, i expected m ≈ interval(0.295, 0.395) to hold.

julia> m
[0.294999, 0.395001]

julia> m.lo ≈ 0.295
true

julia> m.hi ≈ 0.395
true

julia> m ≈ interval(0.295, 0.395)
false

julia> @which m ≈ interval(0.295, 0.395)
isapprox(x::Number, y::Number; atol, rtol, nans, norm) in Base at floatfuncs.jl:278

see also https://github.com/lucaferranti/IntervalLinearAlgebra.jl/blob/main/src/utils.jl#L1

[schillic]

I noticed that this crashes in the latest release. Maybe it is better that way instead of giving the (intuitively) wrong result.

# 0.20
julia> interval(2) ≈ interval(2-1e-10)
false

# 0.21
julia> interval(2) ≈ interval(2-1e-10)
ERROR: promotion of types Int64 and Interval{Float64} failed to change any arguments
Stacktrace:
 [1] error(::String, ::String, ::String)
   @ Base ./error.jl:44
 [2] sametype_error(input::Tuple{Int64, Interval{Float64}})
   @ Base ./promotion.jl:405
 [3] not_sametype(x::Tuple{Int64, Interval{Float64}}, y::Tuple{Int64, Interval{Float64}})
   @ Base ./promotion.jl:399
 [4] promote
   @ ./promotion.jl:382 [inlined]
 [5] max(x::Int64, y::Interval{Float64})
   @ Base ./promotion.jl:459
 [6] isapprox(x::Interval{Float64}, y::Interval{Float64}; atol::Int64, rtol::Int64, nans::Bool, norm::typeof(abs))
   @ Base ./floatfuncs.jl:307
 [7] isapprox(x::Interval{Float64}, y::Interval{Float64})
   @ Base ./floatfuncs.jl:304
 [8] top-level scope
   @ REPL[6]:1