Proper handling of complex numbers and failures

[avik-pal]

Fixes #269

• __findmin ignores NaNs
• TR returns ConvergenceFailure if it stops due to shrinking
• complex numbers gave a complex abstol/reltol
• TR solves an ill conditioned GN step which is not needed.
• Test for an infeasible case!
• Reworks the Termination Conditions and makes everything type stable (if chunksize and such are specified)!
• Return Codes are more meaningful

TODO

• Delete the Manifest before merging
• (Kind-of) Type Stability Fixes for No Chunksize Specified JuliaDiff/SparseDiffTools.jl#271
• Rework the Termination Condition API to be type stable DiffEqBase.jl#954

[codecov]

Codecov Report

Merging #265 (c489b23) into master (1e4c3c0) will not change coverage.
The diff coverage is 0.00%.

@@          Coverage Diff           @@
##           master    #265   +/-   ##
======================================
  Coverage    0.00%   0.00%           
======================================
  Files          19      19           
  Lines        1744    1754   +10     
======================================
- Misses       1744    1754   +10     

Files	Coverage Δ
ext/NonlinearSolveFastLevenbergMarquardtExt.jl	0.00% <ø> (ø)
ext/NonlinearSolveLeastSquaresOptimExt.jl	0.00% <ø> (ø)
src/ad.jl	0.00% <ø> (ø)
src/extension_algs.jl	0.00% <ø> (ø)
src/jacobian.jl	0.00% <0.00%> (ø)
src/default.jl	0.00% <0.00%> (ø)
src/NonlinearSolve.jl	0.00% <0.00%> (ø)
src/raphson.jl	0.00% <0.00%> (ø)
src/pseudotransient.jl	0.00% <0.00%> (ø)
src/levenberg.jl	0.00% <0.00%> (ø)
... and 8 more

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

[avik-pal]

Builds off JuliaDiff/SparseDiffTools.jl#271 & SciML/DiffEqBase.jl#954

[avik-pal]

@ChrisRackauckas this should be good to go