Fix autodifferentiation

[devmotion]

There were some bugs regarding autodifferentiation. This PR also improves results of domain callbacks that rely on a ManifoldProjection.

I had to remove f! from the definition of the Jacobian configuration and replace it with nothing, such that according to the documentation of ForwardDiff

the returned instance can be used with any target function. However, this will reduce ForwardDiff's ability to catch and prevent perturbation confusion (see JuliaDiff/ForwardDiff.jl#83).

[coveralls]

Coverage increased (+8.9%) to 91.129% when pulling e4fe6e6 on devmotion:fix_autodiff into 70ae024 on JuliaDiffEq:master.

[codecov]

Codecov Report

Merging #16 into master will increase coverage by 8.87%.
The diff coverage is 66.66%.

@@            Coverage Diff             @@
##           master      #16      +/-   ##
==========================================
+ Coverage   82.25%   91.12%   +8.87%     
==========================================
  Files           4        4              
  Lines         124      124              
==========================================
+ Hits          102      113      +11     
+ Misses         22       11      -11

Impacted Files	Coverage Δ
src/manifold.jl	88.23% <66.66%> (+32.35%)	⬆️

---

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 70ae024...e4fe6e6. Read the comment docs.