The current version of EstimateZernikesCwfsTask combines the donut stamp generation and zernike estimation into the same task. Because estimating the Zernikes requires donut stamps from each detector in a corner wavefront sensor pair we cannot load this task for each pair from a single dataId. This means that we currently load all the detectors from an exposure and run donut stamp generation and Zernike estimation sequentially for each corner raft. If we split the task into two:

1) Gather the detectors and create the donutStamps . This part is fast and having to do it sequentially is not too much of a bottleneck.

And 2) The task that actually takes the donutStamps and runs the estimateZernikes code. If the donutStampsIntra and donutStampsExtra are just saved under the SW0 sensor dataId for each pair then we could just call this and run Zernike estimation for each pair of donuts in parallel as their own task.

This would speed up the pipeline from its current state.