Different update strategies

[KnutAM]

• Allow not updating jacobian for the first iteration (use the old jacobian from previously converged time step)
• Do not update jacobian when using the SteepestDescent solver

(From FESolvers' perspective, this just implies setting the update_jacobian keyword argument to update_problem!, it is the user that must branch to check what must be updated).

[codecov-commenter]

Codecov Report

Merging #24 (63a4ef3) into main (fa53931) will decrease coverage by 2.31%. The diff coverage is 80.76%.

:exclamation: Your organization is not using the GitHub App Integration. As a result you may experience degraded service beginning May 15th. Please install the Github App Integration for your organization. Read more.

@@            Coverage Diff             @@
##             main      #24      +/-   ##
==========================================
- Coverage   95.12%   92.81%   -2.31%     
==========================================
  Files           7        7              
  Lines         164      181      +17     
==========================================
+ Hits          156      168      +12     
- Misses          8       13       +5     

Impacted Files	Coverage Δ
src/nlsolvers.jl	89.23% <78.94%> (-5.00%)	:arrow_down:
src/QuasiStaticSolver.jl	96.42% <85.71%> (-3.58%)	:arrow_down:

:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more

[KnutAM]

@koehlerson, for the SteepestDescent nlsolver, you expect the following update request, right? update_problem!(problem, Δa; update_residual=true, update_jacobian=false) Is there any case when you don't want this for the SteepestDescent nlsolver, and I should make it optional like I've done for the Newton solver now?

[koehlerson]

I think this should always be true for SteepestDescent solvers. I guess you would group solvers which do request a Jacobian update then into Quasi-Newton methods.

Edit: Sorry for the late reply!

[KnutAM]

Thanks! Yes, but I think it is fine to keep the Newton name, as it is pure Newton-Raphson with default arguments. I'll update the docstring to highlight this!