Bug in device for reshaped views

[ranocha]

Reported by @chriselrod in #156 (comment):

 julia> using ArrayInterface
 
 julia> u_base = randn(10, 10); u_view = view(u_base, 2:3, :); u_reshaped_view = reshape(u_view, 1, length(u_view));
 
 julia> ArrayInterface.device(u_reshaped_view)
 ArrayInterface.CPUPointer()

This u_reshaped_view is not strided, so it should not be a CPUPointer(), but a CPUIndex() (which currently is not supported by LoopVectorization).

[Tokazama]

The core issue here is that we cannot properly dispatch on ReshapedArray (or ReinterpretArray) to a known solution for defines_strides. Previously, we were giving up on strides for both of these if we didn't know it was possible from type information, but recent changes attempt to solve this at runtime.

To fix this we need to either go back to giving up on strides when it isn't known at compile time, get Base Julia to store size information statically, or rework how device is supposed to be used so that it doesn't rely on strides.