A Cube texture is treated at the API level like a 6-element array of 2D textures, so you *should* be able to bind a 2D array view of the cube, and use the Texture2DArray.Load intrinsic.
You'll find that the real trick here is handling the edge cases, since in sampling, if you have a vector that hits right on the edge of the two textures, you need to interpolate across two different textures with potentially different texture coordinates.
Using the same 3-D coordinate system doesn't make sense when you're attempting to fetch a single texel from a cube map (since it's not a 1-1 direct mapping), so we don't expose anything that lets you address single texels on cube textures (you'll notice the offset intrinsics are also not available, partially because the offset direction actually changes depending on which face you're on). Note that if you really just want a non-interpolated value, you can use point sampling.
But if you want to directly address single texels in a cube map, you have to do it in a way that makes sense. This will mean using Texture2DArray, and a lot more work for you if you need to to support 3D directions as addresses.