Group-stepwise ORDC, load reserve provision, and OfflineReserve ORDC supply

[rodrigomha]

Stacked on #235 (base branch rh/reserve_refactor); the diff shows only this feature's commits.

New features (not refactor items) re-expressed on the psy6 reserve tree:

• GroupStepwiseCostReserve: elastic group ORDC. One dense ServiceRequirementVariable per group, a clearing constraint sum(member awards) >= requirement variable, and the group's demand curve (static or time series) priced through the existing delta-PWL path. Mis-paired ServiceModels (GroupReserve with a non-group formulation and vice versa) fail at declaration.
• Load reserve provision under PowerLoadDispatch: the inverse of a generator. Up reserve is committed shed (P - r_up >= 0), down reserve is committed extra consumption (P + r_down <= forecast), gated on an attached service model. A costless load selling reserves errors since nothing pins its consumption.
• OfflineReserve (non-spin) as ORDC supply from storage and loads: non-spinning is upward-only, so it routes like an up reserve everywhere a device supplies it (new UP_RESERVE union). Previously only thermal was wired: the storage reserve-balance multipliers MethodError'd, two storage coverage branches silently skipped the service, complete coverage hit an @assert false, and _modify_device_model! no-opped every OfflineReserve model. The no-op is now scoped to NonSpinningReserve, whose awards ride ReservePowerConstraint instead of the device range expressions.
• Integration test: energy + reserve co-clearing with an elastic service, a stepwise group, and per-resource offers from thermal, storage, and load participants.

Tests live in existing files: group-stepwise testsets in test_services_constructor.jl; market integration, load provision, and the OfflineReserve ORDC scenarios in test_device_reserve_offers.jl; the storage non-spin coverage testset in test_storage_device_models.jl.

Full suite green: 106708/106708. Docs build green.

🤖 Generated with Claude Code

[rodrigomha]

New commit 3479231: OfflineReserve can now supply an ORDC (StepwiseCostReserve) from storage and loads, not just thermal. Non-spin is upward-only, so it routes like an up reserve everywhere (new UP_RESERVE union). This fixes two silent skip branches and an @assert false in the storage coverage constraints, and scopes the _modify_device_model! no-op to NonSpinningReserve only. Full suite 106708/106708.

[jd-lara]

@rodrigomha this PR needs to be rebased

[rodrigomha]

@jd-lara rebased onto main (post #235 squash-merge): the 7 feature commits replayed clean with zero conflicts (git rebase --onto main <235-head>), full suite green on the rebased branch (106708/106708). Base is already main.

[rodrigomha]

Done with the rebase

[github-actions]

Performance Results

Version	Precompile Time
Main	5.911655268
This Branch	5.895921239

Version	Build Time
Main-Build Time Precompile	100.085987451
Main-Build Time Postcompile	5.104126136
This Branch-Build Time Precompile	95.252371564
This Branch-Build Time Postcompile	4.570282716

Version	Solve Time
Main-Solve Time Precompile	250.573936162
Main-Solve Time Postcompile	218.707044414
This Branch-Solve Time Precompile	139.960491362
This Branch-Solve Time Postcompile	89.101204677

[rodrigomha]

CI status after the rebase, fully diagnosed:

• Fixed on this branch (518ae0e, f0fda18): environment resolution. PSY psy6 now depends on the unregistered OpenAPI-models packages, and PSB psy6 on PowerTableDataParser/PowerFlowFileParser - Pkg ignores [sources] of non-root projects, so POM's root/test/docs environments now pin them directly. Docs, build, and comparison jobs are green.
• Remaining test-job failures are upstream PSY psy6 bugs, reproduced and fixed locally (pending review before push): (1) the generated HydroPumpTurbine exporter calls _turbinepump_po, which was never added to the hand-written export helpers; (2) DOCUMENT_PLAN imports HydroPumpTurbine after HydroReservoir, which resolves its unit ids -> unresolved-id error; (3) FuelCurve.fuel_cost::TimeSeriesKey has no OpenAPI serde, which breaks PSB's bundle cache for c_sys5_re_fuel_cost (fixed via the schema's string variant + a document-level key-resolution pass).
• With those PSY fixes applied locally: PSY test_openapi_document 448/448, POM full suite 106708/106708. No further POM commits needed - the pins float on psy6.
• Separate flag: PSB's new bundle cache write is non-atomic; parallel first-builds can read a half-written system.json (UnexpectedEOF) - the old first-serialization race in a new spot. Worth a write-to-temp+rename in PSB.

[rodrigomha]

Restacked on #233 per the network-consolidation sequencing: base is now jd/network_matrix_consolidation and the branch is rebased onto it (4e28ede), so the diff shows only this PR's commits.

• The 7 feature commits replayed clean. Of the two CI environment-pin commits, the PSB transitive-deps one was dropped entirely and the OpenAPI one shrank to docs-env pins only - network matrix consolidation #233 already carries the same root/test pins (same non-root [sources] traps, fixed independently there).
• Note network matrix consolidation #233 pins IOM at jd/network-sources; this stack inherits that.
• Validation on the stacked branch with latest psy6 (which now includes the _turbinepump_po and hydro DOCUMENT_PLAN fixes): compile + zero method ambiguities, and the full suite 106951/106951. The one remaining upstream gap for CI is the FuelCurve.fuel_cost time-series serde in PSY (fix staged locally, pending review) - without it the thermal fuel-cost test fails via PSB's bundle cache.

[luke-kiernan]

Looks fine, no issues at a glance. However, you should probably have others review it too--I'm not very familiar the distinctions between these differennt types of reserves...

[codecov]

Codecov Report

❌ Patch coverage is 88.32487% with 23 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/energy_storage_models/storage_models.jl	80.00%	4 Missing ⚠️
src/static_injector_models/load_constructor.jl	71.42%	4 Missing ⚠️
src/core/reserve_traits.jl	40.00%	3 Missing ⚠️
src/services_models/reserves.jl	62.50%	3 Missing ⚠️
src/services_models/services_constructor.jl	92.68%	3 Missing ⚠️
src/static_injector_models/electric_loads.jl	86.66%	2 Missing ⚠️
...c_injector_models/thermalgeneration_constructor.jl	77.77%	2 Missing ⚠️
src/common_models/add_to_expression.jl	96.15%	1 Missing ⚠️
src/static_injector_models/thermal_generation.jl	96.29%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!