DOC: Summarize best practices

[cluhmann]

Idea or request for content:

I think it would be good to have a set of best practices and/or practices that are "intuitive" but not ideal. The sorts of things I am thinking about might include:

• use of arrays vs. loops when specifying models
• use of pytensor.tensor rather than numpy when dealing with model parameters
• coords rather than shapes
• don't explicitly specify step in pm.sample()
• target_accept=0.99 isn't a fix for poor sampling
• inferenceData objects are canonically named idata rather than trace

and probably many others.

I think the one question is where content like this would live in order to be visible by users. There's nothing else like it in the example gallery. The source of https://www.pymc.io would probably fit, but I'm not sure how much traffic ends up going there. It wouldn't make much sense to write all this up but then not having anyone read it.

[OriolAbril]

Part of that is already covered in https://www.pymc.io/projects/docs/en/latest/contributing/jupyter_style.html#variable-names (and dropdown section). But I agree it would make sense to promote to its own page and extend. It would be a great resource even if only for guidance in writing example notebooks.