Use uniform convention between bytes and str in subprocess outputs
The following discussion from !399 (merged) should be addressed:
-
@clemenci started a discussion: I made this change because I found a problem in a checkout Jenkins job with result objects using not consistently bytes and str. Only later I realized I fixed the same issue with the exact opposite convention in other places (and tests).
Since this change made the test build work I decided to keep it instead of trying to unify the conventions.