Memory loading issue possible workaround

Nobody has claimed this yet.

Hi, i've recently came across the memory loading issue like in the issue ([438](https://github.com/indygreg/PyOxidizer/issues/438)). I am also using pydantic(1.10.2). After I tried around for some time I tried using pydantic from github instead of Pypi. In that case, the error `extension module pydantic cannot be loaded from memory but memory loading required` won't come anymore.

Now the main question is, why there is a difference between using the pre-build wheels from Pypi and using the github repo and building it itself. The problem is, that we want to use pyoxidizer to build independent executables without the requirement to install additional libs (at least for linux).
I've made a [demo repo](https://github.com/henningWoehr/pydantic_test) to easily reproduce this error. There's also a little more documentation on how to reproduce this error.

I hope this is the right place for this issue because I figured, this has to be an issue with how pip is used to installs the wheel, but that's just guess.

Start with the linked demo repository and compare its PyPI pydantic dependency with the build from the GitHub repository, then review issue #438 for the related memory-loading behavior. Done means identifying the wheel or packaging difference and documenting or implementing a path that lets PyOxidizer produce an independent Linux executable.

Written by the indexing model from the issue text.