Working directory is changed after calling libE with threads as communications

[AngelFP]

We discussed this on slack some time ago, but I'll make an issue so that we can better keep track of it.

I noticed that when running with threads, the working directory of the python script is sometimes changed to the workflow_dir_path after the call to libE. I tried this with v1.2 and I don't see a clear way of reproducing the issue. It happens quite randomly, maybe 1 in 10 times running the same script.

So far, I implemented the following workaround in optimas:
https://github.com/optimas-org/optimas/blob/48d42fc4ce79a711625b2b4ba97fea054941c31d/optimas/explorations/base.py#L156

[jlnav]

Hi Angel, sorry for taking a long time to look into this; I'm having trouble myself reproducing it. The workers by design should be changing to the workflow directory, but upon the workers exiting, they should be changing back.

Just to clarify, you're seeing that after the libE call, (all workers should be exited), the current directory stays switched to the workflow directory?