Simpler defaults without FiniteDifferences special cases

[devmotion]

IMO the AbstractFiniteDifference special cases in the default definitions are confusing (see #94...) and lead to code that is less idiomatic and more difficult to optimize.

With increasing dimensionality, avoiding one additional computation of the primal matters less and less, and hence this PR proposes to disentangle the computation of the primal and the jacobian/Hessian/etc. Moreover, also with these changes one can improve performance for individual backends, if possible and desired, by defining a more optimized value_and_jacobian etc.

An additional advantage for higher-order calls value_and_hessian and value_gradient_and_hessian is that it is sufficient to call gradient instead of value_and_gradient.

[codecov]

Codecov Report

Patch coverage: 100.00% and project coverage change: -1.30% ⚠️

Comparison is base (3c18e86) 84.12% compared to head (4c55550) 82.82%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master      #96      +/-   ##
==========================================
- Coverage   84.12%   82.82%   -1.30%     
==========================================
  Files           8        8              
  Lines         466      425      -41     
==========================================
- Hits          392      352      -40     
+ Misses         74       73       -1     

Files Changed	Coverage Δ
src/AbstractDifferentiation.jl	78.18% <100.00%> (-0.28%)	⬇️

... and 3 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[gdalle]

I'm strongly in favor of more simplicity at the expense of a little performance.
I think there might be a conflict with #93 though, gonna try to do the merge

[devmotion]

Ah, missed that there's a merge conflict. I'll fix it.

[devmotion]

The merge conflicts are fixed, the PR is ready for review 🙂

[gdalle]

Looks good to me, apart from the deletion in the @primitive macro, which I don't understand.
Although to be fair I don't understand the macro as a whole. Why do we only have 2 out of 5 (now 4) ..._and_friends functions defined in the package? Where are the rest?