Extensions to MBM and cutting planes techniques.

[dnguyen227]

This PR makes changes to extend MBM and cutting planes techniques.

MBM

• copy_model_with_constraints(::InfiniteModel, ...)
• prepare_max_M_objective(::InfiniteModel, ...) (LessThan + GreaterThan)
• raw_M(sub, ::Vector{<:AbstractJuMPScalar}, ::_MBM)

Cutting Planes

• copy_and_reformulate(::InfiniteModel, ..., ::CuttingPlanes)
• extract_solution(::InfiniteModel)
• add_cut(::InfiniteModel, ...)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.51%. Comparing base (eca54be) to head (cd9dd70).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #135      +/-   ##
==========================================
+ Coverage   99.38%   99.51%   +0.12%     
==========================================
  Files          17       17              
  Lines        1959     2063     +104     
==========================================
+ Hits         1947     2053     +106     
+ Misses         12       10       -2     

☔ 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.

[pulsipher]

See the comments below. Also, we will need some documentation on how these methods are handled for InfiniteModels (there is some nuance).

[pulsipher]

Based on the code I see below, there are a number of gotchas when it comes to copying InfiniteModels, I think we should instead implement JuMP.copy_model(model::InfiniteModel) in InfiniteOpt first to then use here.

[dnguyen227]

Noted, I'll try to get some code going before Monday's meeting and I can try for a PR after we chat.

[pulsipher]

What happens if this is not imported? It seems strange for the user to import a package they do not directly use.

[dnguyen227]

When it's not imported I have all test sets failing (not limited to just raw_M where its called).
MethodError: no method matching InfiniteGDPModel() The function InfiniteGDPModel exists, but no method is defined for this combination of argument types.

[pulsipher]

This is because Julia extensions cannot have additional dependencies not included in the base package. And due to some type piracy in Interpolations, I do not want it to be a core dependency of InfiniteOpt or DisjunctiveProgramming. If you only use it for linear interpolation of M-values, perhaps it makes since to have a lightweight native function instead. It might even make sense just to use constant interpolation.

[dnguyen227]

@pulsipher Tests are failing with MethodError: no method matching copy_model(::InfiniteModel) (passing locally on my clone). Please advise on what I should do with setting compatible versions.

[pulsipher]

@pulsipher Tests are failing with MethodError: no method matching copy_model(::InfiniteModel) (passing locally on my clone). Please advise on what I should do with setting compatible versions.

This is b/c the master version was not be properly set up for testing on CI. However, InfiniteOpt v0.6.2 will be released in about 10 minutes, so just update the Project file to require 0.6.2 and you should be good to go.