Easiest way to do this is to add a parameter for image id number to the current calexp cutout case. If the id number is there, fetch that specific image and then do the cutout. Otherwise fetch an image based on RA/Dec filter. Having both filter and image id in the request should probably be treated as an error.

The current code loads the entire image and then takes a cutout of it. For pixel-level cutouts of id-identified images, I would like to see the Butler's inherent cutout functionality (passing a bounding box into the get call) used instead for efficiency.

For RA/dec locations and boxes, the WCS is required to transform from sky coordinates to pixel coordinates. We could do two I/Os and load the metadata first to get the WCS, then load only the cutout pixels, or we could just load the entire image as is currently done.

See https://confluence.lsstcorp.org/display/DM/Data+Access+meeting+2017-01-30 for the current proposal on how to provide the cutout-from-ID functionality.

The proposal is now captured as endpoints I11 and I12 on the https://confluence.lsstcorp.org/display/DM/API page.

Ready for code review - thanks!

A couple of minor comments, otherwise looks good.

Completed, and deployed to PDAC.