Implement get_dtype and get_device_type for tensors

[jwallwork23]

These, in combination with the existing get_rank, get_shape, and get_device_index, will allow us to create one tensor from another using the same setup.

This will be useful for autograd because the default is to compute the gradient w.r.t. a tensor of ones of the appropriate shape.

We should also make get_device_index a method of torch_tensor, as with get_rank and get_shape.