Add node_id validation

[visr]

Fixes #1780
Fixes #905

As a draft since I still need to review, this is written by Claude.

Plan: Node-ID Consistency Validation via Schema Metadata

Add a node_id_relation trait to Julia schema table types, auto-generate it into Python schema classes, and implement validation in NodeModel that checks node_id consistency across tables.

---

Classification

Relation	Meaning
"equal"	Table's node_ids must exactly match the NodeModel's node_ids (default)
"subset"	Table's node_ids must be a subset
"partition"	All partition tables must be pairwise disjoint, union must equal all node_ids

Partition (static + time pairs): Pump, Outlet, PidControl, LevelBoundary, FlowBoundary, FlowDemand, LevelDemand, UserDemand, TabulatedRatingCurve, Basin

Equal (default): Basin profile/state/concentration/concentration_external/concentration_state, LevelBoundary/FlowBoundary/UserDemand concentration, DiscreteControl variable/condition/logic, ContinuousControl variable/function, ManningResistance static, LinearResistance static

Subset: Basin subgrid/subgrid_time/area, FlowBoundary area, Observation time

Changes Made

Julia — Single Source of Truth (config.jl)

• Added node_id_relation(::Type{<:Table}) function with a default of :equal
• Defined :partition overloads for all Static/Time pairs (Pump, Outlet, LevelBoundary, FlowBoundary, TabulatedRatingCurve, PidControl, UserDemand, LevelDemand, FlowDemand)
• Defined :subset overloads for Basin.Static, Basin.Time (Basin forcing is optional per-node), Basin.Subgrid, Basin.SubgridTime, and Observation.Time
• Exported node_id_relation

Codegen (gen_python.jl, schemas.py.jinja)

• get_models() now includes node_id_relation per table type
• Template emits _node_id_relation: str = "..." class variable on each schema class

Generated (schemas.py)

• All 30+ schema classes now have _node_id_relation with correct values

Non-codegen schemas (area.py)

• Added _node_id_relation = "subset" to BasinAreaSchema and FlowBoundaryAreaSchema

Validation logic (node.py)

• Added _validate_node_ids() method to NodeModel that:
  • Checks "equal" tables: node_ids must match exactly
  • Checks "subset" tables: node_ids must be a subset
  • Checks "partition" tables: must be pairwise disjoint and their union must equal the full set

Wired into model validation (model.py)

• _validate_model() now calls _validate_node_ids() on each node model

Tests (test_validation.py)

• test_node_id_validation_valid — valid model passes
• test_node_id_partition_overlap — node_id in both static and time raises
• test_node_id_partition_missing — incomplete partition raises
• test_node_id_equal_mismatch — missing node_id in equal table raises

Design Note

Basin.Static/Time are :subset rather than :partition because Basin forcing is optional per-node (Julia uses is_complete=false, defaulting to NaN). The disjointness between Basin static and time is still validated by Julia core at runtime.