Necessary clean-ups & possible improvements

[h-vetinari]

Carried over from #43:

There's a fair amount of clean-ups to do here:

• update python
  

  by-the-numbers/environment.yml

  Line 5 in 8ab2886

  - python=3.8

• unpin pandas
  

  by-the-numbers/conda-requirements.txt

  Line 4 in 8ab2886

  pandas<2

• remove cell outputs from what's committed (e.g. using nbstripout git hooks)
• ensure all the notebooks are still runnable, produce useful output (alternatively, remove)
• One other thing I noticed [in count_downloads.py] is that we're doing a lot of pointless downloading - recalculating the download numbers from 2017-2023 is a waste of time, because (now that the .conda stats are in), they should be fully static. We should implement some basic persistence (with an override if we really want to recalculate everything), which would allow us to cut down the downloads to the last few months, rather than currently 8*12=96 (and growing), which takes over half an hour (in binder).