* Add trego configuretion

* Add docstring

* Fix WB2 name

* Fix typos

* Add Trego state part in EgorSTate

* Export trained models from next_points

* Rename ObjData

* Remove grad availability as grad is implemented for all

* Add Trego initial impl

* Refactor trego module

* Fix cobyla cstr tolerance handling

* Rename ObjData

* Add trego activated config

* Move egor config under solver

* Renaming

* Exchange solver and impl

* Move SurrogateBuilder impl in gpmix

* Add multistart in local step

* Fix bug

* Always use slsqp for trego local step

* Compute infill data only once

* Define egor glaobal step

* Use local area sampling

* Fix last best iter computation from best index

* Add trego option in Py binding

* FIx doctests

* Implement refresh_surrogates

* Rename ego_step

* Simplify methods args

* Fix when nlopt selected

* Do not return best_index from trego_step, use new_state

* EGO step return data in new state

* Uniformize state argument position

* Now trego_step use state data only

* One step local/global by iteration

* Increase nb of iterations again

* Increase max_iter for one test

* Relax tolerance for one test

* Refactor to use argmin iteration properly

* Increase max iters in one test

* Fix prev_step_ego and comment

* Comment

* Fix WB2 scaling

* Add traces

* Fix point addition count

* Log trego local step infill criterion max value

* Rename f_min in fmin

* prev_best index is set when best_incex is set

* Fix WB2S criterion

* FIx scale ic usage

* Update prev_best_index with best_index

* Fix fmin computation

* Factorize fmin computation, simplify

* Increase nb of iterations in test

* Update documentation

* Bump 0.21 version

* Fix doc

* Update notebook