Test failures on main

[Robbybp]

See e.g. https://github.com/lanl-ansi/PowerModelsRestoration.jl/actions/runs/11598737054/job/32295250780

I can only reproduce the test-ac-ml-uc-heuristic failure locally. We seem to be hitting an empty array in PowerModels.calc_connected_components.

[noahrhodes]

I assume this update changed something that results in hitting an empty array
lanl-ansi/PowerModels.jl#914

[Robbybp]

Yeah, that's likely the culprit. Have submitted a patch upstream lanl-ansi/PowerModels.jl#933

[Robbybp]

Pulling the most recent PowerModels release, we have two failing tests on MacOS-latest (x86):

test ac redispatch: Test Failed at /Users/runner/work/PowerModelsRestoration.jl/PowerModelsRestoration.jl/test/restoration_redispatch.jl:27
  Expression: isapprox(load_power(result_sim, "1", ["1", "2", "3"]), 9.8492; atol = 1)
   Evaluated: isapprox(7.949205235219295, 9.8492; atol = 1)

test ac redispatch: Test Failed at /Users/runner/work/PowerModelsRestoration.jl/PowerModelsRestoration.jl/test/restoration_redispatch.jl:30
  Expression: gen_power(result_sim, "1", ["3", "4", "5"]) >= 8.5
   Evaluated: 8.000000018656474 >= 8.5

These are testing the result of run_restoration_redispatch(mn_data, PowerModels.ACPPowerModel, nlp_solver). Is the solution to loosen the tolerances, or are the nonunique solutions we're seeing a bug that should be resolved elsewhere?

[noahrhodes]

I suspect it just needs to be relaxed. I assume objective value is still identical, and gen/load increases in consecutive time steps?