IMO, this is a complete waste of time and much of the code doesn't actually read any better afterwards. Some of the mathematical expressions look worse. There may be some merit to splitting the imports but that isn't worth much either, perhaps not enough to warrant trouncing the version control history on those lines.
Another -1 from me. Similar changes to inline examples in the docs have been rejected in the past for the same reasons as Raymond wrote. See issue 4649.
On the other hand, I would be +0 for the tzinfo-examples fixes of the form: foo(x=default) vs. foo(x = default) (This was also the fix suggested on the ML.) Overall, I think it is good to judicially point out to PEP 8 violations where fixes would improve readability, but mechanical reformatting is likely to be net loss. I may fix tzinfo-examples next time I touch it. With addition of timezone class in 3.3, this set of examples needs to be updated.