Consider using DataFrame or xarray internally

[florian-huber]

Currently the data is stored in COO form with

• .row & .column for the 2D indexing
• .data as a numpy structured array which essentially is a table, able to contain different data types.

This functionality is also provided by pandas DataFrames or xarray Datasets which would bring more functionalities and make the code more readible. Main questions are:

• Effort worth it?
• How about performance?

[florian-huber]

This is (or was) tested in #4 .
While this makes part of the code more accessible, performance is much better when using numpy structured arrays. In particular slicing was much faster.