WIP: Array function

[elisno]

(I'm quite new to the whole package development process, so any feedback would be appreciated.)

I've modified existing methods in weeks.jl to evaluate the ILT of vector valued functions. This avoids repeated function evaluations when sampling the function (before the FFT).
It might be a good idea to make a seperate method, since scalar functions now output a vector of
length = 1.

Currently, calling w.(t::AbstractArray) isn't optimized and it outputs a vector of incorrectly nested vectors. Apparently, the length of the vector-function has to be an integer power of 2, otherwise the ordering of the output will be off (probably due to the call to fftshift).

I've added a test case which is similar to the complex one. It only evaluates w(t), for a single value of t.

I haven't tried it yet, but this method could be generalized any array valued functions, so long that the function samples FF are stored along the first dimension (plans for the FFT would then be always be along the "columns").

[jlapeyre]

Thanks for the comments and PR. It could be that writing methods for broadcast! is the way to go. I'll be able to look at in a week or so.