Check stderr for subprocess call errors

[jlstevens]

This PR simply moves #59 from a fork to this repo so that I can attempt to fix the tests.

[jlstevens]

TODO

• Add comment explaining issue with dodgy return code as detailed in param issue.

[jbednar]

I'm happy to have the actual code change from this PR or #59, but the updates to the tests rely on a slew of other libraries that aren't appropriate to include in the autover codebase, even as optional tests. If it's possible to boil those tests down to a small amount of standard Python, it would be great to include them, but otherwise I'd say the probability that these tests will cause maintenance issues for us over time is much higher than the probability that they will catch actual errors. I'd strongly favor moving towards a design like #66 where released versions are no longer ever calling any subprocesses at all, at which point no such tests should be necessary in any case.

[jlstevens]

If it's possible to boil those tests down to a small amount of standard Python, it would be great to include them, but otherwise I'd say the probability that these tests will cause maintenance issues for us over time is much higher than the probability that they will catch actual errors.

I agree and I've updated the PR to just include just the fix (they are still in the history and shown to be passing, thanks @julioasoto!). As you say, once #66 is working this code path should be mostly moot (for release users at least).

@jbednar Think this simplified PR can now be merged?