Handle downstream usage issues

[ChrisRackauckas]

It seems this library wasn't properly tested...

[codecov]

Codecov Report

Merging #88 (a9a82b6) into master (6347ae5) will increase coverage by 37.22%.
The diff coverage is 58.33%.

@@             Coverage Diff             @@
##           master      #88       +/-   ##
===========================================
+ Coverage    0.00%   37.22%   +37.22%     
===========================================
  Files           8        5        -3     
  Lines         459      231      -228     
===========================================
+ Hits            0       86       +86     
+ Misses        459      145      -314     

Impacted Files	Coverage Δ
src/utils.jl	12.38% <12.90%> (+12.38%)	⬆️
src/raphson.jl	50.00% <60.86%> (+50.00%)	⬆️
src/NonlinearSolve.jl	100.00% <100.00%> (ø)
src/ad.jl	100.00% <100.00%> (ø)
src/jacobian.jl	43.24% <100.00%> (+43.24%)	⬆️

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

[andreasnoack]

The tests here are not running at all. The code coverage has been zero for more than half a year https://app.codecov.io/gh/SciML/NonlinearSolve.jl. The reason is that you only run the tests for

NonlinearSolve.jl/test/runtests.jl

Line 10 in 22ed463

if GROUP == "All" || GROUP == "Interface"

but GROUP never has any of those values

NonlinearSolve.jl/.github/workflows/CI.yml

Lines 15 to 16 in 22ed463

	- Core
	- Downstream

. Might be good to add an else branch that errors.

[ChrisRackauckas]

Finally.

[andreasnoack]

Can the downstream failure safely be ignored?

[ChrisRackauckas]

Yes, this is getting a v1.0 release, and MTK is moving to SimpleNonlinerSolve.jl. It took awhile because this library had to be split to get the dependencies correct, which is a very breaking change. It also unifies the arguments (abstol instead of tol), so it's a very breaking update. Basically, it's finally updated from being an overlooked intern project to a full compliant piece of SciML, with all of the breakage required to make it conform to the rules of the interface. The update should be relatively sane though since most things would actually just move to SimpleNonlinearSolve.jl: the "bigger" Newton here is really just for big systems which isn't used all that much downstream.