fix(python): Fix interchange protocol data buffer dtype

[stinodego]

I believe there is an issue with the way data buffer dtypes are represented across various implementations of the protocol (ours, pyarrow, pandas, ...).

The issue is that the data buffer is assigned the column dtype. This is only correct for integers and floats. Categoricals, strings, and datetime types have a some integer as their physical representation. The data buffer should have this physical data type associated with it.

The fix in this PR is simple enough, but this cannot be merged until other libraries fix their from_dataframe implementation. I opened some issues:

• pandas: BUG: Interchange object data buffer has the wrong dtype / from_dataframe incorrect pandas-dev/pandas#54781
• pyarrow: [Python] Interchange object data buffer has the wrong dtype / from_dataframe incorrect apache/arrow#37598

When those libraries update their from_dataframe implementations per the issues above, the roundtrip tests should pass and this can be merged.

[stinodego]

PyArrow and pandas have updated their implementations. This can now be merged. But let's give it a bit of time for integration's sake. I will merge this when we start merging breaking changes for 1.0.0.

[stinodego]

I'm going ahead with this one as well. I think it makes sense to make this change together with the new from_dataframe implementation. That way, any integration issues can be addressed within the same Polars update.