Delta formulation and free starts

[manuelma]

This PR introduces the delta formulation for the representative periods model. The main changes are as follows:

• When generating time-slices for represented blocks, we ignore the resolution parameter and we instead use the resolution of the representative_periods_mapping parameter. This is because we only need time-slices at mapping resolution to write the deltas.
• For represented periods, the only constraints are the min_node_state and node_state_capacity constraints for nodes with is_long_term_storage=true. No other constraint is generated outside of representative periods.
• In the above two constraints, the node_state is obtained from a JuMP expression corresponding to the delta equation.
• We don't generate any variables for represented periods.

Also we introduce temporal blocks with free start using middle histories, superseding #1254 (now closed).

Fixes #1233

Checklist before merging

• Documentation is up-to-date
• Unit tests have been added/updated accordingly
• improved type stability or, when outside the scope of the pull request, at least indicated where and how the code is not properly typed
• Code has been formatted according to SpineOpt's style
• Unit tests pass

[codecov]

Codecov Report

❌ Patch coverage is 92.50000% with 15 lines in your changes missing coverage. Please review.
✅ Project coverage is 86.33%. Comparing base (6f7ee3d) to head (1151d61).
⚠️ Report is 1 commits behind head on master.

Files with missing lines	Patch %	Lines
src/data_structure/temporal_structure.jl	91.77%	13 Missing ⚠️
src/constraints/constraint_cyclic_node_state.jl	85.71%	1 Missing ⚠️
src/variables/variable_node_state.jl	88.88%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1253      +/-   ##
==========================================
- Coverage   86.48%   86.33%   -0.15%     
==========================================
  Files         148      148              
  Lines        4660     4735      +75     
==========================================
+ Hits         4030     4088      +58     
- Misses        630      647      +17     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[DillonJ]

Great stuff. @manuelma reports the following build times for a version of our SpineH2IRL model:

Before any improvements: 45 min
Improved implementation (PR #1247): 7 min
Delta formulation (This PR): 3 min

@datejada @aporrasc : Can you guys test on the WP5 test model?

[datejada]

Great stuff. @manuelma reports the following build times for a version of our SpineH2IRL model:

Before any improvements: 45 min Improved implementation (PR #1247): 7 min Delta formulation (This PR): 3 min

@datejada @aporrasc : Can you guys test on the WP5 test model?

It looks promising; it's the 12RP case from @aporrasc that I have, which goes from around 10m to 4.5m to create the model. So, kudos for that. Some open questions from my side are:

1. The objective function should be the same when solving (or close to it). I didn't check that so far.
2. How does this scale with more RPs (50RP, 100RPs...)

Thanks!