Refactor architecture to use Tables (DuckDB)

[abelsiqueira]

Description

Given the changes that are happening in the input (ref #289) and that happened in the model, it makes sense to use Tables to store most of data, instead of converting the data into other structures only to convert them back to dataframes in the model.

We should benchmark and evaluate the changes to make sure they are good ideas.

Sub issues

• Separate the reading and processing in the IO #544
• Return tables (a.k.a. DuckDB) or dataframes as the output format #115
• Refactor the inputs to use DuckDB as much as possible #706
• Change representative_periods to use DuckDB #713
• Change base_periods to use DuckDB
• Change constraints_partitions to use DuckDB
• Change graph to a mix of Graph and use DuckDB
• Use dataframes instead of dict for RP weight matrix #495
• Change construct_dataframes to use tables (a.k.a. DuckDB) or input data frames #611

[abelsiqueira]

We should update or remote this in favour of the discussion happening in #688

[datejada]

Closing this issue in favour of Epic #701, which has all the relevant issues that are not finished here (at the moment of closing this issue).