toNdArray does not support string tensor

[arkanoid87]

@[1,2,3,4,5].toTensor.toNdArray # OK

@["a","b","c"].toTensor.toNdArray # KO

Error: type mismatch: got <Tensor[system.string]>
but expected one of:
func toUnsafeView[T: KnownSupportsCopyMem](t: Tensor[T];
    aligned: static bool = true): ptr UncheckedArray[T]
  first type mismatch at position: 1
  required type for t: Tensor[toUnsafeView.T]
  but expression 't' is of type: Tensor[system.string]
proc toUnsafeView[T](ndArray: NumpyArray[T]): ptr UncheckedArray[T]
  first type mismatch at position: 1
  required type for ndArray: NumpyArray[toUnsafeView.T]
  but expression 't' is of type: Tensor[system.string]

[Clonkk]

Indeed, the current implementation focuses on numerical data, char, bool, int*, uint*, float, complex.

Problem with Tensor[string] is that string are gc-allocated in Nim which means the Tensor is not going to be contiguous so you cannot copyMem it - therefore all you can do is loop over the datas - which is not efficient.

Tensor[string] is really just storing string with fancy indexing on top of it.