Remove DataFrames completely

[abelsiqueira]

Now that DuckDB is being used almost everywhere, we can finish the job and remove DataFrames.
This mostly means:

• TulipaVariables (and friends) field indices is now a DuckDB thing. To loop over it we don't need to use for row in eachrow(indices), just for row in indices.
• Some things are harder to do, since they must use SQL now. Namely, subsets.
• There should be some memory and speed improvement.

It should also be possible to do the following improvements, which can be considered "nice to have":

• Don't even store indices, instead store connection and when it is desirable to loop, call the query to create the looping object. I'm not sure this is actually necessary or useful, because I believe the DuckDB indices will never be store materialised. So in practice it should be the same efficiency/memory use.
• Make loops look nicer by creating an API to loop over the rows and the container/expressions at the same time. For instance, instead of doing

  [... for (row, expr1, expr2) in zip(cons.indices, cons.expressions[:expr1], cons.expressions[:expr2])]
  

  we can create a function to wrap that:

  [... for (row, expr1, expr2) in loop_over(cons, with_expr[:expr1, :expr2])
  

  Some thought is necessary to ensure (some) type stability.

Related issues

Related to #701

Checklist

• I am following the contributing guidelines
• Tests are passing
• Lint workflow is passing
• Docs were updated and workflow is passing

[github-actions]

Benchmark Results

	68c3bcc...	f74f0d5...	68c3bcc... / f74f0d5...
energy_problem/create_model	36 ± 2 s	35.7 ± 3.5 s	1.01
energy_problem/input_and_constructor	27.8 ± 0.17 s	27.7 ± 0.068 s	1.01
time_to_load	2.62 ± 0.023 s	2.66 ± 0.035 s	0.987

	68c3bcc...	f74f0d5...	68c3bcc... / f74f0d5...
energy_problem/create_model	0.245 G allocs: 11.9 GB	0.243 G allocs: 12.1 GB	0.98
energy_problem/input_and_constructor	30.8 M allocs: 1.18 GB	26.8 M allocs: 0.952 GB	1.24
time_to_load	0.159 k allocs: 11.2 kB	0.159 k allocs: 11.2 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).

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.86%. Comparing base (68c3bcc) to head (f74f0d5).
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1035   +/-   ##
=======================================
  Coverage   97.86%   97.86%           
=======================================
  Files          29       29           
  Lines         982      985    +3     
=======================================
+ Hits          961      964    +3     
  Misses         21       21           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[datejada]

This comment is for my understanding: ::Int is here to ensure type stability since we know in advance that the result will be an Int, but Julia can't know it in advance because it comes from a SQL query. Is that correct?

[abelsiqueira]

Exactly. The draft PR #1055 will add more of these.

[datejada]

Thanks for the PR. The code looks cleaner, and deleting dependencies is good 😄 The code was also initially review on the Tulipa working day. I left a comment for my learning porpuse, so that it is why it is approved now.