Synchronize and prevent repeated package caching in all presets

[HannesWell]

With this PR I suggest to store pythonFile computed in the cpython.presets.python.cachePackage() in a static variable and to return it if it is already computed to prevent caching of the python packages multiple times.
I assume the result should be the same in each call?

This method is called twice when using the embeddedpython.Python class (in a stripped variant) and each call takes at least one third of the 2 seconds initialization time of embeddedpython.Python. So also caching the path would speed up the initialization significantly.

Actually this method could be synchronized to prevent concurrent caching of the packages. If you agree I can update this PR accordingly.

[saudet]

Right, that and synchronizing it make it more consistent with how Loader.load() works. Sounds like a good idea! While we're at it, could you do the same for numpy, scipy, tensorflow, gym, opencv, llvm, and tvm as well?

[HannesWell]

Right, that and synchronizing it make it more consistent with how Loader.load() works. Sounds like a good idea! While we're at it, could you do the same for numpy, scipy, tensorflow, gym, opencv, llvm, and tvm as well?

Great.
Sure. I have applied it to the other mentioned presets and also searched for other cachePackage() methods but found none. So I think those are all.