Python 2 has unicode literals and utf-8 available if you declare them at the top of your file. You don't need Python 3 for that.
Unicode isn't a "fancy thing" that is best left to "people that know what they're doing." It's the difference between displaying a name such as Siån correctly and as a bunch of hieroglyphics.
Besides, even for experienced developers who know what they're doing, if you're imposing any form of ceremony or special considerations around using Unicode, you're increasing the risk of bugs and mistakes.
That's why I say Unicode should be made the default. It's legacy encodings such as Latin-1 that should be treated on a need-to-know basis.
Unicode may be of little relevance to some disciplines, such as scientific computing for example, but in others, such as modern web development, it is a deal breaker.
I also don't need to be told that defaults are important. I know they're important.
But it is still a minor improvement under the scope of a ~6 year migration that has been absolute hell to everyone involved. Under any other circumstance, I'd be right there with you saying It's A Really Good Thing, but if we have to pay this high of a cost for it, then the improvement here looks pretty meh to me.
If you want to insist on evaluating this on some absolute scale irrespective of its cost, then I don't want anything to do with that.
Finally, one might argue that whether this is actually minor or not is irrelevant. What's relevant is that a boatload of people perceive the delta between Python 2 and Python 3 to be incredibly small, and yet, the amount of work to migrate is dauntingly large. There's a discrepancy there regardless of whether you disagree with others' valuations of the improvements in Python 3.