RFR (S) 8232168: Fix non wide char canonicalization on Windows

[Langer, Christoph]

Hi Alan,

> I have this on my list to look at. I think we'll need to figure out a
> path to separate the usages (the JPLIS agent usage is technical debt, we
> should have addressed that issue a long time ago).

I agree that it's a good thing to explore how the different usages of calls to canonicalize could be disentangled. However, can't we defer that to another item? This one seems like a straightforward bugfix to me, with some cleanup attached. And I've prepared some more cleanup in that area already which is dependent on this fix - so I'd appreciate if this wouldn't be procrastinated too much :) 

Thanks
Christoph