It has for 2 years now. 3.6 was the tipping point, such a fantastic release.

See

https://www.jetbrains.com/research/devecosystem-2018/python/

For 2018 stats

It's already been since 2015. Every major Python dependency has been Python3 compatible for several years, and many have already gone Python3 only.

If you have a good test suite it's pretty easy to migrate, you can easily do a couple thousand lines per day. There were a few third-party testing tools that used to be fairly popular that are no longer well maintained, so if your tests are written in those then you might be looking at re-writing the app from scratch. But otherwise it's fairly straightforward to port.

Even if you did your string encoding just using guess-and-check, which most people did, it still doesn't take all that long to upgrade as long as there are tests.

I would say yes. I'm pretty sure most people are writing all new projects exclusively in python3, except if you need them to run on old distros without any dependencies (lots of distributions shipped for years with python2 but not python3 installed by default). The ecosystem of 3 is very mature - I've been running my servers' production code for 5+ years in python3 and had no major issues with that choice. The only thing left is "legacy codebases".

Python 2 EOL on January 1, 2020 is pushing a lot of folks to seriously look at this. This is a big thing we're looking at right now at Khan Academy.