numpy
Remove reinforcement learning notebook from executable build
As noted elsewhere the atari-py package on which the reinforcement learning tutorial was based was deprecated (#87), and the project that replaces it EULA(s) that need to be reviewed and may no longer be suitable for open source.
We had been using the deprecated atari-py, but this has many problems, including the requirement of building wheels locally, and now dependency issues (see #143).
Given all this, I think it's past time that we remove these dependencies, which will necessitate switching the RL notebook to no longer be executable. I took the simplest possible approach to this in this PR. Ultimately, I'd like to start a discussion about what to do with content that is not executable, but I'll leave that for another issue/meeting.
