Foolproof sitepackages_dir_cmd, and raise error when it fails

[ashwinvis]

Fixes #209

[ghost]

I ran into this today. I'd really like to see this fix applied, since it seems to break my current workflow.

[pfmoore]

@ashwinvis Are you still interested in this? The appveyor tests are failing, which should be looked at.

[ashwinvis]

I can update the PR. Seems like #214 is fixing the CI, so I can rebase the branch if it will be merged. Sounds good?

[pfmoore]

Sure, if that works I'm fine with it. The error in question is "ImportError: No module named pew._utils" which seems a bit more fundamental, though (although I don't see how it relates to this PR).

[pfmoore]

FWIW, I just tested master on my local (Windows) PC, and I see the test failures that #214 is addressing, but not the pew._utils one. But I was testing with Python 3.8, not Python 2.7, so maybe it's something Python 2 specific.

Ultimately, I'm happy to approve this PR if the tests pass, but I can't make the decision to accept a PR with test failures, sorry.

[ashwinvis]

Can you retrigger the Appveyor CI - it could be something to do with their Python build / environment a year ago, which hopefully got fixed now.

[pfmoore]

The only way I have of doing that is to close & reopen the PR, which I'll do. I don't have admin access to the Appveyor setup.

[pfmoore]

Sigh, that looks even worse 🙁

I don't have much more time to spend on this right now. Looks like there's some not-insignificant bitrot to address here :-( Probably something to pick up first on #218...