Single implementation of column type autodetection

[eroell]

Description of feature

Column type autodetection (e.g. whether numeric or categorical) is required at different stages of ehrapy

E.g. the autodetect option of encoding here relies on the adata.var[EHRAPY_TYPE_KEY] tag.

This is set when

• reading a dataframe here
• moving something from obs to X here.

We might want to harmonize that? It is the df.select_dtypes("number").columns pattern pretty consistently it appears at a first glance, but seems to be scattered.

Other places where this occurs and should be harmonized:

• in the CohortTracker

TODOs:

• collect where this type detection is done
• move it to a single implementation

[eroell]

tableone might have something fairly robust we could be interested in too

[Lilly-May]

I've merged this issue with the related issues #649 and #637 into the new issue #701 for a better overview. Therefore, I'll close this issue here.