Refactor io into reading and processing and create table tree structure

[abelsiqueira]

The function to create graph et al from a csv folder has been split into two functions. The first reads the csv folder into a new TableTree structure. The second processes the TableTree structure into graph et al.

Pull request details

Describe the changes made in this pull request

List of related issues or pull requests

Closes #544

Collaboration confirmation

As a contributor I confirm

• I read and followed the instructions in README.dev.md
• The documentation is up to date with the changes introduced in this Pull Request (or NA)
• Tests are passing
• Lint is passing

[github-actions]

Benchmark Results

	49a243d...	4b606a5...	49a243d.../4b606a57a4e1e1...
energy_problem/create_model	0.106 ± 0.0055 h	0.106 ± 0.005 h	0.996
energy_problem/input_and_constructor	26.9 ± 0.37 s	27.3 ± 0.42 s	0.987
time_to_load	2.56 ± 0.012 s	2.55 ± 0.0054 s	1

	49a243d...	4b606a5...	49a243d.../4b606a57a4e1e1...
energy_problem/create_model	0.749 G allocs: 31.2 GB	0.749 G allocs: 31.2 GB	1
energy_problem/input_and_constructor	0.249 G allocs: 4.34 GB	0.249 G allocs: 4.34 GB	1
time_to_load	0.153 k allocs: 14.5 kB	0.153 k allocs: 14.5 kB	1

Benchmark Plots

A plot of the benchmark results have been uploaded as an artifact to the workflow run for this PR.
Go to "Actions"->"Benchmark a pull request"->[the most recent run]->"Artifacts" (at the bottom).

[suvayu]

This all looks pretty straight forward to me. I only have one comment. This still assumes everything is read from CSV, e.g, create_input_dataframes_from_csv_folder directly creates the table_tree. It would be nice if you could split this into:

flowchart LR
    A[file] -->B(DataFrame)
    subgraph "in Julia (TEM)"
    B --> C(TableTree)
    end

Loading

Then TIO can easily replace "reading from file"

flowchart LR
    A[TIO] -->B(DataFrame)
    subgraph TEM
    B --> C(TableTree)
    end

Loading

[abelsiqueira]

Thanks for the review. It's true, eventually we'll have to split more, but it will also depend on TulipaClustering.jl, so I'm ignoring it for now.

[suvayu]

The longer this is delayed, longer I've to code without a clear target. It's one of the reasons reading ESDL has been slow.