Recover from PIRLS failure by returning finitial

[palday]

One potential downside to this approach is that it does make it possible to get stuck and/or have the optimizer believe that the initial values are the optimum. The former should be clear from (lack of) progress. I guess we could check that m.optsum.finitial != objective(m) and error if it does. The downside is that we have try in the objective, which we can assume is a fairly hot function.

• add tests (achieved by fitting a model that previously didn't work with fast=false)
• add entry in NEWS.md
• after opening this PR, add a reference and run docs/NEWS-update.jl to update the cross-references.
• I've bumped the version appropriately

[codecov]

Codecov Report

Patch coverage: 100.00% and no project coverage change.

Comparison is base (c99a2d3) 95.82% compared to head (8cbaadf) 95.83%.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #616   +/-   ##
=======================================
  Coverage   95.82%   95.83%           
=======================================
  Files          35       35           
  Lines        3259     3263    +4     
=======================================
+ Hits         3123     3127    +4     
  Misses        136      136           

Flag	Coverage Δ
current	95.83% <100.00%> (?)
minimum	95.72% <100.00%> (+<0.01%)	⬆️
nightly	95.83% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files Changed	Coverage Δ
src/generalizedlinearmixedmodel.jl	90.47% <100.00%> (+0.11%)	⬆️

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.