DSEM implementation in FIMS

[e-perl-NOAA]

Describe the problem, idea, or request that you have and, if possible, provide a solution.

We would like to add internal capacity of dsem into FIMS similar to how rceattle does it (thank you @grantdadams) but in TMB and not rTMB.

As myself and Fernando (what I have affectionately named my Copilot/Gemini) work on this with @Andrea-Havron-NOAA's guidance (because I have absolutely no idea what im doing, I just have copilot , a bit of time 🕐, and a dream ☁️), I am now going back and actually adding an issue and sub-issues for this task because there should be an issue before there is a PR ... ooops (see PR #1594)

The basic plan is:

1. Add gmrf function in distributions
2. Add precision builder assemblage of functions that initially just include dsem but have the capacity to include others (spatial, etc.). The precision builder needs to be able to handle covariance.
3. Link get_observed and get_expected in gmrf function to different parameters/derived quantities in the model this will happen in the information.hpp file similar to SetupData() and will need a info->variable_map in rcpp_dsem.hpp - this will be harder to do because we have to link up the x vector with multiple parameters and how to get the ordering correct for this.
4. Add gtests for gmrf and precision builder
5. Create dsem rcpp file
6. Create front end R dsem functionality
7. Create tests and vignettes for dsem in FIMS

Additional context

No response

Which OS are you seeing the problem on?

No response

Which version of this package are you seeing the problem on?

No response

[grantdadams]

Nice to see this scoped out, @e-perl-NOAA! 🙂 Gentle prioritization note — DSEM (with issues #1656–#1658) is still R&D and not operational for us, so putting it in the Parking lot feels right. From the CIE review, I'd suggest focusing on the core operational features of FIMS first (sex/seasonality/growth, reference points, projections) and avoiding feature bloat. This doesn't seem critical to operational assessments, but a good one to revisit once we figure out best practices and the key operational features of FIMS are implemented and have users.