We get data out of p4python as strs, but we weren't treating them as such.

This change adds some logic to decode changesets before processing them, which

should improve the situation.

Ran unit tests.