Cropped ISIS Cube and ALE Generated ISD

When running the following pipeline, I am getting unexpected results. Talking with @acpaquette these results are sensical because of how ISIS implements cropping.

Here is the pipeline:
- Generate an ISIS cube
- Crop the cube and propagate SPICE
- Use `isd_generate` to create an ISD from the cropped cube

My expected behavior is that the ISD generated by `isd_generate` will be appropriate to use with the cropped ISIS cube. By appropriate, I mean that a `campt` call and some knoten sensor model function will return the same results, plus or minus the accepted tolerance. In reality, the results are quite different. Why?

ISIS tracks the offsets in the cropped cube and carries around the full observation SPICE. Therefore, the ISD created by ALE is the full observation SPICE. But, the cropped cube is just that, a cropped image. The CSM has no idea that the ISD is provided for an uncropped image.

The result is that a test to ensure that say, pixel (10,10) in the cropped ISIS image returns the same lon/lat between both `campt` and `knoten` requires that I add whatever the start sample/line used for the `crop` command. That information is not recorded in the ISD. It is available via `cathist`.

I do not know that I have a suggested fix. Options that come to mind (without a lot of thought about the trickle down effects):
- "Fix" ISIS so that a cropped cube crops the SPICE as well. (I say "fix" because I think ISIS is working as intended.)
- Fix `isd_generate` to create a proper cropped ISD if the input ISIS cube is cropped