update `requirements-write.txt`

[drbenvincent]

The current requirements file is out of date and potentially too minimal

https://github.com/pymc-devs/pymc-examples/blob/dce63a03b894d39a0a39a4773f9a0226078391c1/requirements-write.txt#L1-L9

Todo

• [ ] Replace pymc3 with pymc, possibly with a >=4.0 version tag.
• [ ] We don't need to have total coverages of all packages used in the repo, but it would be helpful to add more commonly used packages such as:
  • matplotlib
  • numpy
  • xarray
  • seaborn
  • bambi
  • jupyterlab-code-formatter, isort (useful when writing notebooks in the browser, rather than VS Code for example)
  • patsy
  • scipy
  • daft (for making causal DAG's)
• [ ] Check version on pre-commit is still appropriate
• [ ] Update instructions in final section of the contributing guide to include the instructions on how to build a pymc-examples-env environment. E.g. conda create --name pymc-examples-env --file requirements-write.txt

Happy to take feedback on this of course.

[OriolAbril]

I agree with most of the points, but would try to keep the requirements there minimal.

Expanding and being explicit on all downstream dependencies we use explicitly is a good idea I think too. Therefore, numpy, scipy, pandas, xarray, matplotlib, arviz. I would also add graphviz (unless there is an option for pymc[all] or something of the sort that installs pymc and optional dependencies. Regarding the others I think we could add some extra libraries that are not pymc dependencies (direct or indirect) but I would be quite strict on that. I would clearly not include patsy for example which I think is only used in one example, whereas seaborn or bambi are used much more extensively and should probably have a place there. Formatting things are fine with me as long as the docs are clear that it is not required to used them, the only condition is getting pre-commit to pass