NMPG solver

[aldma]

This PR intends to add a simple solver for regularized problems min f+g based on the classical proximal-gradient method. It can use nonmonotone globalization and spectral (or BB) stepsizes.

[dpo]

At least in principle, those solvers are already implemented as part of R2DH. If the nonmonotone strategy isn't yet implemented there, it should be.

[codecov]

Codecov Report

❌ Patch coverage is 86.26374% with 25 lines in your changes missing coverage. Please review.
✅ Project coverage is 84.05%. Comparing base (e0f214d) to head (733cc24).
⚠️ Report is 287 commits behind head on master.

Files with missing lines	Patch %	Lines
src/NMPG.jl	84.90%	24 Missing ⚠️
src/R2DH.jl	95.65%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           master     #337       +/-   ##
===========================================
+ Coverage   61.53%   84.05%   +22.52%     
===========================================
  Files          11       13        +2     
  Lines        1292     1819      +527     
===========================================
+ Hits          795     1529      +734     
+ Misses        497      290      -207     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[dpo]

Things to benchmark later to better understand the difference between NMPG and R2DH:

• Armijo acceptance condition vs $\rho_k$ (Armijo requires local Lipschitz gradient)
• stopping conditions
• non-monotonicity with average formula vs max.