Long answer: A big factor was that new devs already knew Git because it had taken off spectacularly in the open source world. In only a few years SVN was legacy and everyone switched to Git (and Github) for open source.
Windows shops that had been using SVN had a very natural transition to Git, with lots of tooling and experience available.
But there were indeed a lot of Windows shops that used other, mostly commercial and centralized version control systems before. While they might have had some advantages over CVS and SVN or when used for enterprise, those advantages disappeared when compared to the new distributed version control systems like Mercurial and Git.
Even though Git is not exactly easy to learn, it was still vastly better than the old version control systems. In many ways the old systems just really sucked in comparison.
As for Git Bash: You could always install Git through Cygwin, but I don't remember Windows people using the command line much. There were many Git GUI applications for Windows, and IDEs provided integrations as well. E.g. Eclipse came with a completely commandline-free Git client and integrated UI that worked flawlessly on Windows.
As for Mercurial vs Git: it wasn't completely clear from the start that Git would "win", as they are very similar. I think network effects, Github, and the Linux kernel halo effect are mostly responsible for the difference in use.