Add toArray1D<type>, toArray2D<type> etc.

I have been trying to incorporate diffsharp into an existing codebase which is based on Math.NET Numerics but I am having trouble with getting standard vectors and matrices from diffsharp tensors. My intention is to check correctness of current calculations with diffsharp.

let t1 = dsharp.tensor [ 0.0 .. 0.2 .. 1.0 ]
let a = t1.toArray()

This returns System.Array which is incompatible with 'a [ ]for some reason.

Is there a way to get typed F# arrays (or 2d arrays) from tensors? A good example to what I need would be that tensorflow can return ndarray from tensors.

[dsyme]

It returns a System.Array, which can then be downcast to the right kind of array

The exact type of the array is the natural .NET corresponding type for the element type of tensor, seen here for example: https://github.com/DiffSharp/DiffSharp/blob/c1d7eb613539cae3ae11a7cca4df230503ef84e0/src/DiffSharp.Backends.Torch/Torch.RawTensor.fs#L222

If we add .toArray1D<float>(), toArray2D<int64>() etc. I guess they will do the job of casting to the requested element type before delivering the results.

[gbaydin]

Implemented in #360