Tangi Migot

Hey guys! Is there something blocking here?

Oh, I see. Thanks @BoZenKhaa !

or we make a giant TODO and add simple examples in the tests

I believe this is very similar to `performance_profile` in [SolverBenchmark.jl](https://github.com/JuliaSmoothOptimizers/SolverBenchmark.jl/blob/b1c51f2a80512c5bc3e72455733d63e7e8b44dd7/src/profiles.jl#L21), where the input is `stats::Dict{Symbol,DataFrame}` whose keys are solver names and values data frames from the solver call.

Hi @ojwoodford Very sorry for the delay, we are currently understaffed. I think that would be a good Idea, feel free to open a PR with a proposition. As much...

> Here are a few minor comments. ~I don’t understand why the parameters should be inside the solver object?!~ (nevermind, I see). It could actually, but we would need a...

Everything regarding the constraints on the domain of the parameters should be modeled in the parameter set, the package SolverParameters.jl has a bunch of functionalities for this (even open sets)....

> Wouldn't constraints be modeled directly in the `BBNLPModel`? Mainly yes, but some minimal constraints are also handled here. Beyond the optimization of parameters, it is also good to make...

> I noticed that `trunk` and `tron` interfaces do not allow to pass parameters to the sub-solvers (e.g. `increase_threshold` parameter for `TRONTrustRegion`), and don't have them as keyword arguments. With...

@dpo @d-monnet I finally updated this PR! The default parameters are now centralized at the beginning of each file, and include default value documentation so we would no longer need...