Only compute Jacobian in RLM if last step success

[Affie]

The estimate lms.p is only updated on a successful step, therefore a Jacobian is not needed if the last step failed. This PR skips updating the Jacobian if the last step failed.

[codecov[bot]]

Codecov Report

Merging #303 (6d29ba4) into master (c4bc49d) will increase coverage by 0.01%. The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #303      +/-   ##
==========================================
+ Coverage   99.74%   99.76%   +0.01%     
==========================================
  Files          77       77              
  Lines        7087     7089       +2     
==========================================
+ Hits         7069     7072       +3     
+ Misses         18       17       -1     

Files	Coverage Δ
src/plans/nonlinear_least_squares_plan.jl	100.00% <ø> (ø)
src/solvers/LevenbergMarquardt.jl	100.00% <100.00%> (ø)

... and 1 file with indirect coverage changes

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

[Affie]

No, tagging is not urgent as I can work on the master branch locally and CI will still pass as it's only a performance enhancement.

[kellertuer]

Ok, nice, I'll merge it to master and it will be part of the next release.