Trying out different ways of extending Pandas according to the documentation:
https://pandas.pydata.org/docs/development/extending.html
Extending dtypes is quite complex, and therefore I have note handled it to completion yet. Some work may be found under extension_types, but will be a work in progress.
But its a very good place to start, and a good way of exploring what next step to take and organize by.
Placed under "pandas_object"-folder, this adds methods to all "pandas-objects", including Series... This can get messy fast, and is to be avoided.
If you have frequently repeated dataset-types with strongly coupled behaviour-needs, this might be a good option, but it might be unclear by name and type, that, really, this is only a slightly modified pandas-dataframe...
With introducing your own namespace under DataFrames and Series, you can make them do new things. But this would probably be better with "bigger domains" than multiple Subclasses would? Or maybe having many isnt a problem?