Sb-arch-opt

- Also update cumulative pop when not storing results
- Added logo
- Constrain numpy to 1.x for now

- Add SEGOMOE ask-tell interface and pymoo Algorithm
- Updated to pymoo 0.6.1

- Update to Trieste 2.0
- Add OSS license checks for dependencies
- Updated SEGOMOE integration

- Improved hierarchical sampling by randomizing group selection
- Display population statistics (non-failed, feasible, optimal points)
- Do not use problem-specific correction by default
- Fixed bug with problems needing correction of continuous variables and using an auto corrector
- Added correction ratio metric to problem statistics
- Fixed bug with ArchSBO running on a problem with an explicit design space
- Added interface to the [Egor](https://github.com/relf/egobox) optimizer (#11)
- Switch to random forest classifier as default hidden constraint strategy for ArchSBO

Test problems:
- Added bi-objective version of the realistic turbofan test problem
- Added surrogate model version of the simple turbofan test problem
- Added multi-stage rocket design problem
- Added mixed-discrete versions of the GNC problem

Updates:
- Implemented problem-agnostic corrector algorithms to use if `all_discrete_x` is available
- Switched to LHS as fallback sampler
- Bug fixes

ArchSBO-specific updates:
- Updated default settings for hidden constraint strategies: support KPLS and auto-scale random forest
- Updated default infill selection: only use ensembles
- Added gradient-based infill optimization after the evolutionary optimization
- Updated to SMT 2.2
- Do not force cont relax kernel for PLS for SMT>=2.2
- Added more logging

- Fixed instantiation of continuously-relaxed mixed-discrete Kriging models when using PLS
- Support categorical hierarchical design variables for Kriging models (SMT)
- Support changing DOE size after the fact for doe_algo
- More open interface for storing/loading (intermediate) results; including support for non file-based operations