Provide an option to exclude some AD backends

[amontoison]

close #324
cc @jbcaillau @PierreMartinon
It will be easier for you to exclude some ADbackend with the option exclude_backend.

using ADNLPModels, NLPModels

f(x) = (x[1] - 1)^2
T = Float64
x0 = T[-1.2; 1.0]
lvar, uvar = zeros(T, 2), ones(T, 2)
lcon, ucon = -T[0.5], T[0.5]
c!(cx, x) = begin
  cx[1] = x[2]
  return cx
end
nlp = ADNLPModel!(f, x0, lvar, uvar, c!, lcon, ucon, backend = :optimized,
                  excluded_backend=[:jprod_backend, :jtprod_backend, :hprod_backend, :ghjvprod_backend])

ADNLPModel - Model with automatic differentiation backend ADModelBackend{
  ReverseDiffADGradient,
  EmptyADbackend,
  EmptyADbackend,
  EmptyADbackend,
  SparseADJacobian,
  SparseReverseADHessian,
  EmptyADbackend,
}

[github-actions]

Package name	latest	stable
ExpressionTreeForge
JSOSuite
PartiallySeparableNLPModels
PartiallySeparableSolvers
SolverTest

[tmigot]

The current way to achieve this is to set the mentioned backend to EmptyADbackend, but if you guys think that might be useful why not.

[tmigot]

Looks overall good, thanks @amontoison !
Could you also update the section https://jso.dev/ADNLPModels.jl/dev/performance/ for the doc and maybe add a very short unit test?

[amontoison]

The current way to achieve this is to set the mentioned backend to EmptyADbackend, but if you guys think that might be useful why not.

I forgot that we can do that!
I found it in the documentation: https://jso.dev/ADNLPModels.jl/dev/performance/#Use-only-the-needed-operations

I don't think this PR will bring any advantage. I will just update the documentation to specify the backends needed for Ipopt / Knitro / MadNLP.

using ADNLPModels, NLPModels

f(x) = (x[1] - 1)^2
T = Float64
x0 = T[-1.2; 1.0]
lvar, uvar = zeros(T, 2), ones(T, 2)
lcon, ucon = -T[0.5], T[0.5]
c!(cx, x) = begin
  cx[1] = x[2]
  return cx
end
nlp = ADNLPModel!(f, x0, lvar, uvar, c!, lcon, ucon, backend = :optimized,
                         jprod_backend = ADNLPModels.EmptyADbackend,
                         jtprod_backend = ADNLPModels.EmptyADbackend,
                         hprod_backend = ADNLPModels.EmptyADbackend,
                         ghjvprod_backend = ADNLPModels.EmptyADbackend)

[amontoison]

I updated the documentation in #328.