feat: strengthen results comparator foundation

docs/RESULTS_SCHEMA.md

@@ -64,6 +64,8 @@ Attach these attributes to `/`:
 | `unit_system_note` | string | no | User-provided self-consistent unit note |
 | `dof_convention` | string | yes | `UX,UY,UZ,RX,RY,RZ` |
 | `sign_convention` | string | yes | `Abaqus-compatible` |
+| `precision` | string | yes for Phase 1 in-memory result model | `double` |
+| `index_type` | string | yes for Phase 1 in-memory result model | `int64` |
 
 ## `/metadata`
 Recommended datasets or attributes:
@@ -189,7 +191,7 @@ The final component labels should be cross-checked with the accepted Abaqus refe
 
 ## Phase 1 Mandatory Outputs
 The first complete linear static solver path must write:
-- root metadata attributes: `schema_name`, `schema_version`, `solver_name`, `dof_convention`, `sign_convention`.
+- root metadata attributes: `schema_name`, `schema_version`, `solver_name`, `dof_convention`, `sign_convention`, `precision`, and `index_type`.
 - `/model/nodes/ids`
 - `/model/nodes/coordinates`
 - `/model/elements/ids`
@@ -293,6 +295,8 @@ Required CSV columns:
 Rules:
 - CSV files are reference inputs for tests, not the primary FESA result storage format.
 - The comparator must preserve node-label matching and must not rely on row order alone.
+- The comparator must require FESA `U` component labels `UX`, `UY`, `UZ`, `RX`, `RY`, `RZ` with `position = NODAL`, `entity_type = node`, and `basis = GLOBAL`.
+- Duplicate FESA output node ids, duplicate CSV node labels, missing FESA nodes, missing CSV columns, and nonnumeric CSV values are comparison failures.
 - The comparison report may be stored under `/referenceComparison`.
 - Reaction CSV, stress CSV, or section force CSV formats must be documented before automated use.