Improve cumulative_integral

[devmotion]

Related to #389, this PR optimizes cumulative_integral by returning nothing for non-numeric types. The PR also improves allocations for numeric types a bit.

Fixes #385.

On the master branch:

julia> A = ConstantInterpolation([3.1, 2.5, 4.7], [0.0, 0.25, 0.75])

julia> DataInterpolations.cumulative_integral(A, false)
Float64[]

julia> DataInterpolations.cumulative_integral(A, true)
2-element Vector{Float64}:
 0.775
 2.025

julia> f(A, cache_parameters) = @allocations DataInterpolations.cumulative_integral(A, cache_parameters)

julia> f(A, false)
1

julia> f(A, true)
8

julia> using Unitful

julia> A = ConstantInterpolation(rand(5)u"m", (1:5)u"s");

julia> DataInterpolations.integral(A, 4u"s")
ERROR: ArgumentError: zero(Quantity{Float64}) not defined.
...

With this PR:

julia> A = ConstantInterpolation([3.1, 2.5, 4.7], [0.0, 0.25, 0.75])

julia> DataInterpolations.cumulative_integral(A, false)
Float64[]

julia> DataInterpolations.cumulative_integral(A, true)
2-element Vector{Float64}:
 0.775
 2.025

julia> f(A, cache_parameters) = @allocations DataInterpolations.cumulative_integral(A, cache_parameters)
f (generic function with 2 methods)

julia> f(A, false)
1

julia> f(A, true)
2

julia> using Unitful

julia> A = ConstantInterpolation(rand(5)u"m", (1:5)u"s");

julia> DataInterpolations.integral(A, 4u"s")
1.549774483763152 m s

[Ickaser]

Awesome, thanks @devmotion !