New documentation and tutorials structure

[Zethson]

Description of feature

Our tutorials suffer from at least the following issues:

1. They are too long and verbose.
2. They do not suggest best practices or which of the many options to use.
3. They do not explain how to work with time series data
4. They are hard to navigate because of a weird sidebar structure and the sphinx gallery.

Therefore, I suggest that we:

1. Remove the sphinx gallery and just use text links. A very smart friend of mine told me years ago that this is the way to go.
2. Use lots of dropdowns to explain details but not make the tutorials seem too long.
3. Use lots of see also sections to refer to other parts of the documentation and especially the API. Instead of showing 5 survival analysis functions in one big tutorial, we should show off one and then refer to the API for the other options.
4. Use a more clear structure for the tutorials like:

What we really need is a complete ehrdata example dataset which has every slot populated. No matter how, even if we just add some random stuff to the physionet one. @eroell

-10 minutes to ehrapy (anndata, ehrdata, scanpy, ehrapy) which replaces and merges the current introduction to ehrapy and MIMIC-II IAC if possible. Also introduces the readers and file formats.

• Working with time series data (lots of work needed to get this ready)
• Working with text data (medcat and ontology mapping)
• Survival analysis
• Trajectory inference
• Causal analysis (causal discovery (NA) & causal inference)
• Bias detection (merges with cohort tracking)
• Machine learning (simple model, deep learning & time series, dataloaders & sampling)
• Using ehrapy with large datasets