Drop cppoptlib

[zfergus]

I did this, but I have mixed feelings about it.

Pros:

• The stop criteria and logic are now fully contained and should make PolySolve's Solver easier to understand
• We are using an outdated version of cppoptlib, so these changes should help avoid downstream conflicts (Is there some special reason to use cppoptlib from 2018 instead of the current one? #63)

Cons:

• It ended up being almost a copy-paste from cppoptlib (older version) as we use a lot of features in Problem but little in Solver
• It still requires cppoptlib for two line-search methods: CppOptArmijo and MoreThuente

We can talk about whether to merge these changes here. If not, we can keep the improved comments in Problem.

[codecov]

Codecov Report

Attention: 54 lines in your changes are missing coverage. Please review.

Comparison is base (811c9a7) 89.00% compared to head (ca0611e) 87.27%.

Files	Patch %	Lines
src/polysolve/nonlinear/Criteria.cpp	35.71%	45 Missing ⚠️
src/polysolve/nonlinear/Solver.cpp	91.35%	7 Missing ⚠️
src/polysolve/nonlinear/Problem.hpp	60.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #66      +/-   ##
==========================================
- Coverage   89.00%   87.27%   -1.73%     
==========================================
  Files          51       48       -3     
  Lines        2001     2028      +27     
==========================================
- Hits         1781     1770      -11     
- Misses        220      258      +38     

Flag	Coverage Δ
polysolve	87.27% <68.23%> (-1.73%)	⬇️

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

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

[teseoch]

I never liked the criteria class, because we couldn't extend it. To me, that is the only thing you copied. I would clean and refactor it a bit, so in the end, it is not a copy paste.

Regarding the two LS strategies, we can do without or integrate them in the new framework. Note that they don't properly work with the step valid stuff