Pass-through AbstractOptimizerAttribute

[klamike]

Trying to use nonlinear_diff_model then set_optimizer_attribute:

julia> using MadNLP, MadNLPGPU, CUDA, DiffOpt, JuMP

julia> m = DiffOpt.nonlinear_diff_model(MadNLP.Optimizer)
A JuMP Model
├ mode: DIRECT
├ solver: MadNLP
├ objective_sense: FEASIBILITY_SENSE
├ num_variables: 0
├ num_constraints: 0
└ Names registered in the model: none

julia> set_optimizer_attribute(m, "array_type", CuArray)
ERROR: MathOptInterface.UnsupportedAttribute{MathOptInterface.RawOptimizerAttribute}: Attribute MathOptInterface.RawOptimizerAttribute("array_type") is not supported by the model.
Stacktrace:
 [1] throw_set_error_fallback(model::DiffOpt.Optimizer{MathOptInterface.Utilities.CachingOptimizer{…}}, attr::MathOptInterface.RawOptimizerAttribute, value::Type; error_if_supported::MathOptInterface.SetAttributeNotAllowed{MathOptInterface.RawOptimizerAttribute})
   @ MathOptInterface julia_depot/packages/MathOptInterface/Q3V1z/src/attributes.jl:662
 [2] throw_set_error_fallback(model::DiffOpt.Optimizer{MathOptInterface.Utilities.CachingOptimizer{MathOptInterface.Bridges.LazyBridgeOptimizer{…}, MathOptInterface.Utilities.UniversalFallback{…}}}, attr::MathOptInterface.RawOptimizerAttribute, value::Type)
   @ MathOptInterface julia_depot/packages/MathOptInterface/Q3V1z/src/attributes.jl:653
 [3] set(model::DiffOpt.Optimizer{MathOptInterface.Utilities.CachingOptimizer{MathOptInterface.Bridges.LazyBridgeOptimizer{…}, MathOptInterface.Utilities.UniversalFallback{…}}}, attr::MathOptInterface.RawOptimizerAttribute, args::Type)
   @ MathOptInterface julia_depot/packages/MathOptInterface/Q3V1z/src/attributes.jl:626
 [4] set(m::Model, attr::MathOptInterface.RawOptimizerAttribute, value::Type)
   @ JuMP julia_depot/packages/JuMP/7eD71/src/optimizer_interface.jl:1264
 [5] set_attribute
   @ julia_depot/packages/JuMP/7eD71/src/optimizer_interface.jl:1517 [inlined]
 [6] set_attribute
   @ julia_depot/packages/JuMP/7eD71/src/optimizer_interface.jl:1526 [inlined]
 [7] set_optimizer_attribute(model::Model, attr::String, value::Type)
   @ JuMP julia_depot/packages/JuMP/7eD71/src/optimizer_interface.jl:77
 [8] top-level scope
   @ REPL[5]:1
Some type information was truncated. Use `show(err)` to see complete types.

julia> function MOI.set(model::DiffOpt.Optimizer, attr::MOI.AbstractOptimizerAttribute, value)
           MOI.set(model.optimizer, attr, value)
       end

julia> set_optimizer_attribute(m, "array_type", CuArray)

julia> @variable m x
x

julia> @constraint m c1 x ≥ 1
c1 : x ≥ 1

julia> @objective m Min x^2
x²

julia> optimize!(m)
The following options are ignored: 
 - array_type
This is MadNLP version v0.8.12, running with cuDSS v0.7.1
...

(Note the "The following options are ignored" message is a bug in MadNLP master that will be fixed before the next release, as you can see below it is in fact using cuDSS)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.14%. Comparing base (dc86850) to head (71e701a).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #335      +/-   ##
==========================================
+ Coverage   89.13%   89.14%   +0.01%     
==========================================
  Files          16       16              
  Lines        1961     1963       +2     
==========================================
+ Hits         1748     1750       +2     
  Misses        213      213              

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[andrewrosemberg]

I'm trying to think if there is any attribute that could break diffopt if it got passed but I don't think so. LGTM after format and could be nice to add some tests.