The Data Science Handbook is not explicitly about coding practices, but it deliberately includes a large amount of software engineering and lengthy example scripts. Much moreso than you will find in a typical data science book, so you might want to take a look and see if it's a good fit for your needs.
Full disclosure: I wrote that book.
I think this is all excellent advice. The only thing I would add is that it's good to pick up a book o analytics or DS to give you a broad overview of the subject and some practical tips. Personally I learned from Data Analysis with Open Source Tools. I wrote a book on data science that was aimed at people like you. And I haven't read Data Science from Scratch, but I know the author and I expect that it is quite good too.
Definitely 3. I recently published a data science book that used Python 2, and in retrospect I wish I had used python 3. Both are in widespread use, and there are plenty of things that only support 2 (example: Amazon Redshift). I think 2 is still more common, but 3 is the way of the future.