Ensure that a unique problem module exists before importing

[benzwick]

This fixes unexpected behaviour that occurs if a problem module exists in more than one location or if a problem module throws an exception during import.

Previously, there were two things that could go wrong:

1. The problem module is defined in both the examples directory and the current directory in which case the user might be unaware that the example problem takes precedence over the one in the current directory.

2. If the module exists but raises an exception during import then that exception is caught and an attempt is made to import the module from the current directory instead. Further exceptions are then caught and a RuntimeError is raised which (misleadingly) informs the user that the module could not be found.

This made it difficult to debug example problems because exceptions raised during import were caught by the exception handler.

References:

• https://docs.python.org/3/library/importlib.html#checking-if-a-module-can-be-imported

[benzwick]

Before merging this @mikaem let me just check what happens if a problem module exists somewhere else on Python's path.

[benzwick]

The previous commit didn't consider the case where a problem module is found on Python's path elsewhere than the problems directory or the current directory. Now, instead of checking if multiple problem modules are found the origin of the module is printed before it is imported.

[benzwick]

Should I squash this down into a single commit and create a new pull request @mikaem ?

My first attempt at improving the problem module import behaviour was a bit too complicated and doesn't work for all cases. The second one is much simpler but doesn't actually do what is stated in this pull request description. It simply checks if the module exists before attempting to import it. This way exceptions raised during the loading of the module are handled separately to the exception raised if a module is not found.

[mikaem]

Thanks for all the fixes @benzwick:-)