Stop swallowing exceptions

[Qqwy]

Fixes #57

This change ensure that rather calling PyErr.print() which prints the exception and then forgets about it, we call PyErr.restore() which restores it as 'current' exception.

If the calling Rust code:

• Returns a (), this will now trigger the same exception once we reach Python
• Uses PyErr::take to check for a current exception, it will fetch the exception and is able to manipulate it and turn it into a PyResult in whichever way it sees fit.
• Doesn't use PyErr::take to defuse the exception and returns something other than () itself, then Python will turn the originally thrown exception into a SystemError with the message SystemError: <method 'example' of 'builtins.YourCustomClass' objects> returned a result with an exception set. The original exception is included as cause of this exception.

[vorner]

Looks nice, thank you :-).

Just one question. Did you try what happens if both there's a pre-existing exception and one is raised during the logging? Will it overwrite it, or will there be some kind of „Multiple exceptions“ thing.

[Qqwy]

The current implementation will keep the pre-existing exception intact, i. e. "first exception wins".

[Qqwy]

I ran cargo fmt now to ensure the linting workflow passes.

The other workflows seem to not be specific to this branch; they seem to be failing when fetching Python versions and the like.