refactor: split the pandas code required for converting results to DataFrames out of the pandas backend

[datapythonista]

I think there is a difference between:

• A pandas backend that is able to operate a pandas dataframe based on Ibis expressions
• The functionality to convert a database cursor from another backend into a pandas dataframe

But currently, some code of the latter lives in the pandas backend. This creates problems, as the whole pandas backend is required to be loaded for any other backend to work.

[cpcloud]

I think this would be good to get into 3.x and help us move away from pandas as required dependency.

[jreback]

i am. it sure why we care if pandas is a required dep

[cpcloud]

i am. it sure why we care if pandas is a required dep

Because we want to move away from it being the primary way we deal with in-memory results.

[cpcloud]

At the very least, there needs to be an alternative way to get back something that isn't a DataFrame.

[jreback]

sure not averse to that but need a fully workable alternative first

[cpcloud]

Closing this as stale.