ExaModels.Optimizer fails when variables have bounds #199
Description
julia> m = Model(() -> ExaModels.Optimizer(madnlp, CUDABackend()))
A JuMP Model
├ solver: madnlp running with ExaModels
├ objective_sense: FEASIBILITY_SENSE
├ num_variables: 0
├ num_constraints: 0
└ Names registered in the model: none
julia> @variable m x ≥ 1
x
julia> @objective m Min x^2 + x + 1
x² + x + 1
julia> optimize!(m)
ERROR: KeyError: key MathOptInterface.ConstraintIndex{MathOptInterface.VariableIndex, MathOptInterface.GreaterThan{Float64}}(1) not found
Stacktrace:The reason is we skip adding variable bounds to con_to_idx. Note this is not an issue with ExaModel(m) since that path doesn't create the IndexMap