rhl requests: "During your cleanup, can you fix the run methods to accept unpacked data (i.e. no DataRef s)? Add runDataRef to do the unpacking".
And: "We need to move all the functional code out of the CmdLineTask s. So e.g. ReduceArcTask should either not be a CmdLineTask, or it needs to defer to a "real" task. The exact level of splitting is TBD."

All Task s should return a Struct in preference to a single value.

Fix IdentifyLinesTask._DefaultName.

lineList should be lineListFilename.

hassan, I'm sending this to you as reviewer because the changes are so extensive (several thousands of lines) I don't know how it should be reviewed. I suggest that, rather than one person looking at every line changed, any stakeholders (including rhl and cloomis) have a look at revised C++ APIs, and then someone looks over the python changes. Or we could just say that this is what we want to work off going forward, I merge and any complaints are resolved as future tickets.
I've verified that the pipeline produces similar results before and after this change (I say "similar" rather than "identical" because there are some tiny algorithmic fixes to fitting rejection somewhere that had a small, apparently positive effect; there's also a change to the wavelength fitting in its own commit, but I'm not including that in the comparison).

I haven't yet updated the integration test, but I want to get the code moving towards merge.

For drp_stella, GitHub says "5,448 additions and 7,429 deletions."

Ok let's see if @rhl and @cloomis at least have time this week to review. I think our outstanding unit test discussion is useful for the next time, as such tests are very useful tools to help minimize reviews by individuals.

Merged to master after obtaining consensus on Slack.