Support ROI tables that don't start at 0,0,0

[jluethi]

Currently, the parsing shifts ROIs to start at 0, 0, 0. We will have cases where this isn't possible anymore or hasn't been done. e.g. using external OME-Zarr files with ROIs, processing ROIs during multiplexing (when ROIs may shift due to registration), maybe changes during stitching of images within wells.

Additionally, when we start using ROIs for e.g. Organoids, they will not necessarily start at 0, 0, 0. Not an issue for illumination correction, but let's not make the 0, 0, 0 assumptions for e.g. labeling.

For now, the 0, 0, 0 is a good start. See #8 (comment) for details.

Let's use this issue to track that we may need to change this, e.g. when implementing registration for multiplexing.