LMCMA does not respect lower upper boundaries

[Yneth]

LMCMA does not use lower and upper boundaries in any way, which results in sampling of the invalid states of X.
May apply to other CMA ES algorithms

[Evolutionary-Intelligence]

Currently, PyPop7 only supports unconstrained optimization. We are planning to design a new library to support constrained optimization based on PyPop7.

[Yneth]

I've crafted a small POC on making the LMCMA ask-tell-based. It is not perfect right now, but I really like the approach of the Tell/Ask API, as it can be easily extended to many other things via composition (boundary handling or repeated logic like budget counting/stop events) and can be used in distributed environments.

If you'd like, I can share a small script of how it looks.