IPNewton() sparsity support

[schrimpf]

This makes IPNewton() compatible with alternate types (e.g. sparse) of Hessians and Jacobians. It addresses one part of #609

This is a work in progress. Some remaining issues:

• The passing of array type for the Jacobian of the constraints is ugly. I made it an argument to IPNewton, but it would be better as a field of TwiceDifferentiableConstraints, just like H is a field of TwiceDifferentiable. That requires changing NLSolversBase
• Add en example of a sparse problem to the documentation
• Ensure IPNewton() benefits from sparsity. The changes make IPNewton() work with a sparse Hessian, but I haven't examined whether its appropriately preserved and exploited throughout the algorithm. Should also bench mark the performance versus dense matrices.
• Improve tests. The only test is a simple equality constrained problem. It is not organized the same way as tests for other aspects of the package.

[pkofod]

Appreciate it! Let me know when it's ready for review.

[codecov]

Codecov Report

Merging #977 (24a38a8) into master (cdebfca) will increase coverage by 1.31%.
The diff coverage is 88.88%.

❗ Current head 24a38a8 differs from pull request most recent head f5835da. Consider uploading reports for the commit f5835da to get more accurate results

@@            Coverage Diff             @@
##           master     #977      +/-   ##
==========================================
+ Coverage   83.92%   85.23%   +1.31%     
==========================================
  Files          42       42              
  Lines        3066     3184     +118     
==========================================
+ Hits         2573     2714     +141     
+ Misses        493      470      -23     

Impacted Files	Coverage Δ
...tivariate/solvers/constrained/ipnewton/interior.jl	76.44% <0.00%> (-0.58%)	⬇️
...tivariate/solvers/constrained/ipnewton/ipnewton.jl	96.17% <87.50%> (-0.34%)	⬇️
...ariate/solvers/second_order/newton_trust_region.jl	97.07% <100.00%> (+0.66%)	⬆️
src/types.jl	84.15% <100.00%> (+36.38%)	⬆️
src/multivariate/solvers/first_order/bfgs.jl	90.66% <0.00%> (-1.12%)	⬇️
src/univariate/printing.jl	17.64% <0.00%> (-1.11%)	⬇️
src/multivariate/optimize/interface.jl	81.13% <0.00%> (-0.60%)	⬇️
...c/multivariate/solvers/zeroth_order/nelder_mead.jl	84.84% <0.00%> (-0.51%)	⬇️
src/multivariate/solvers/constrained/fminbox.jl	79.25% <0.00%> (-0.08%)	⬇️
... and 17 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cdebfca...f5835da. Read the comment docs.

[Vaibhavdixit02]

@schrimpf do you plan to continue this?