The current situation is actually not that bad. With my students, the topic is quickly explained and mastered. But it's not a common occurrence.
However, it is true that it's still not as easy as it could be. Some reasons.
- Python is older than java.
- Python is shipped as part of MacOS and Linux, but using the system Python can cause issues, so it's a trap.
- Python deals with a lot of compiled extensions. Scipy embeds C, fortran and assembly. It's a hard problem.
- The Python project has *tremendously* less resources than java.
- Politics. Lots of it.
- Debian et Red Had packagers made a mess by splitting the python setup in many small packages.
- Python 2/3 transition took attention span from the problem during a decade.
- You often install many python versions on the same machine, which makes things harder.
- Python is used by a huge number of non programmers. Much more than java. This shows in the reported difficulties.
- There are way too many ways to install python. So many combinations of failure modes.
- Microsoft screw up with their store not packaging "py" or the core dev by not packaging it on Unix.
- Anaconda brings as many problems as it solves, and is popular because of enterprise limitations.