Add user control of oldctlfile in profiles

[iantaylor-NOAA]

The profile wrapper is great, but for petrale we're having an issue of convergence which I think is related to the sequence of counting down from the estimated value then counting up, all the while using the control.ss_new for the oldctlfile input, which means that the initial values for the non-dev parameters in each step in the profile are the estimated values from the previous step. In general this is a good thing as it improves estimation, but it's a problem when there's a big jump in the parameter.

For a petrale M profile, {nwfscDiag} created the vector below, and estimation with M = 0.142675 starting from the estimated values from the previous M = 0.09 step led to non-converged estimates that did not match the base model with that same M.
[1] 0.140000 0.130000 0.120000 0.110000 0.100000 0.090000 0.142675 0.150000 0.160000 0.170000 0.180000 0.190000

In the profile_control branch I have added user control of the oldctlfile input so that we can change to something like petrale_control.ss for all steps in the profile, rather than the control from the previous run.

A few alternative approaches to this problem worth thinking about (and/or splitting into a separate issue) are:

• run separate profiles, first counting down from the estimated value, then counting up, where both would use oldctlfile = "control.ss_new". This would remove the discontinuity that is messing up petrale profiles, and
• use the original model with estimated parameter rather than re-running the profile with the parameter fixed at the estimated value so save a little time. Maybe the re-run was driven by the
  need for matching settings related to the prior (issue inclusion of priors in likelihood? #6).

[chantelwetzel-noaa]

Thanks @iantaylor-NOAA. I distinctly remember doing profiles for petrale sole in 2019 was particularly difficult with many runs not converging. I like your second suggestion and think that this could be done by the package fairly easily. While I definitely think that finding an approach that produced converged runs reliably is the goal, I want to bring to your attention the rerun_profile_vals function if you were not aware. This function reruns specific values and then does all the processing and plotting again. This function has an input of run_num which corresponds to the run number for the non-converged run that can be found in the csv with "results" at the end of the name.

[iantaylor-NOAA]

@chantelwetzel-noaa, thank you for the tip on rerun_profile_vals(). I remember the discussions when you created that function but forgot that it existed. Running all the petrale models from the same starting values instead of control.ss_new did not magically fix the non-convergence of a few models in each profile so I will give re-running a try.