In cc1e91fc90, running the crosstalk task was moved above assembleCcd, which broke the crosstalk correction in ip_isr.  Due to the fact that HSC uses a custom ISR this fact was not visible in ci_hsc, and in fact no tests were checking for this error.  The tests of the crosstalk task itself were unaffected by changes in isrTask.  The problem only showed up when we started processing lsstCamera data (ts8 and in particular phosim) through obs_lsstCam.

I propose that we revert this change.  It was made to support DECam, but I do not understand why it was necessary so I think it's better to fix obs_decam than ip_isr, but would like to understand why the DECam code was written this way before taking a final decision.  In discussion, Meredith pointed out that she didn't want to have to assemble CCDs to process inter-CCD crosstalk, but I don't think that this would be necessary (you'd use the Detector to iterate over the amplifier segments, applying suitable flips as described in the Detector's amplifier objects)