Split OutlierDetectionStep into multiple steps

[stscijgbot-jp]

Issue JP-3682 was created on JIRA by Brett Graham:

The outlier detection step currently handles several different "modes" in sometimes very different ways:

• imaging
• spec
• coron
• tso
• ifu

The docs mention that the parameters for the step are different depending on the mode:

https://jwst-pipeline.readthedocs.io/en/latest/jwst/outlier_detection/arguments.html

The step spec includes many items that are entirely unused in some modes (which is not fully documented).

These modes also expect very different input types and produce different results.

Different modes use different file suffixes, none of which are currently handled by the shared suffix machinery.

It seems reasonable that these should be split into different steps. However this will require:

• new pars- references for the new steps
• replacing pars-outlierdetectionstep

[stscijgbot-jp]

Comment by Ned Molter on JIRA:

I started looking into what is needed here on the cal side.  It looks like something we'd definitely like to do eventually, and I made a quick draft of what it might look like, which can be found in [this draft pull request|[https://github.com//pull/8722].]  However I sort of "jumped the gun" as this needs discussion with INS teams and would require parameter reference file changes, so assigning it to David again for input.

[stscijgbot-jp]

Comment by David Law on JIRA:

Initial discussion at the Sep 4 JP meeting; will be considered as part of possible 11.2 work.

[stscijgbot-jp]

Comment by David Law on JIRA:

Now that https://jira.stsci.edu/browse/JP-3347 is resolving without the need to use the IFU algorithm on FS/MOS data (which would have confused this effort significantly) I think we can unblock this.  I know at the same time work has been going on to reorganize the documentation, per https://jira.stsci.edu/browse/JP-3342

Tyler Pauly , pushing this back over to you.  Is this still desirable from your side based on the recent doc updates?

[stscijgbot-jp]

Comment by Ned Molter on JIRA:

IMO this is still worth doing, although the documentation updates make it less urgent/crucial.  I'm going to resurrect my (now ancient) PR for this and see if the changes do indeed look helpful to folks.